WO2010025356A2 - System and methods for reducing perceptual device optimization time - Google Patents

System and methods for reducing perceptual device optimization time Download PDF

Info

Publication number
WO2010025356A2
WO2010025356A2 PCT/US2009/055348 US2009055348W WO2010025356A2 WO 2010025356 A2 WO2010025356 A2 WO 2010025356A2 US 2009055348 W US2009055348 W US 2009055348W WO 2010025356 A2 WO2010025356 A2 WO 2010025356A2
Authority
WO
WIPO (PCT)
Prior art keywords
subject
feature
stimuli
features
stimulus
Prior art date
Application number
PCT/US2009/055348
Other languages
French (fr)
Other versions
WO2010025356A3 (en
Inventor
Bonny Banerjee
Lee S. Krause
Alice E. Holmes
Rahul Shrivastav
Original Assignee
University Of Florida Research Foundation, Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US12/201,598 external-priority patent/US9319812B2/en
Priority claimed from US12/201,492 external-priority patent/US9844326B2/en
Application filed by University Of Florida Research Foundation, Inc. filed Critical University Of Florida Research Foundation, Inc.
Publication of WO2010025356A2 publication Critical patent/WO2010025356A2/en
Publication of WO2010025356A3 publication Critical patent/WO2010025356A3/en

Links

Classifications

    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/72Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235Details of waveform analysis
    • A61B5/7264Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/12Audiometering
    • A61B5/121Audiometering evaluating hearing capacity
    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04RLOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R2225/00Details of deaf aids covered by H04R25/00, not provided for in any of its subgroups
    • H04R2225/43Signal processing in hearing aids to enhance the speech intelligibility
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04RLOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R25/00Deaf-aid sets, i.e. electro-acoustic or electro-mechanical hearing aids; Electric tinnitus maskers providing an auditory perception
    • H04R25/70Adaptation of deaf aid to hearing loss, e.g. initial electronic fitting

Definitions

  • the present invention is related to the field of hearing testing, and more particularly, to techniques for more efficiently assessing hearing capabilities and tuning hearing-enhancement and audio devices.
  • Subject testing arises in a wide variety of contexts and ranges from posing a battery of test questions to eliciting a subject's response to different types of stimuli.
  • a prime example of the latter type of testing is the technique used to fit a hearing-impaired subject, or patient, with a cochlear implant system.
  • a suitable speech coding and mapping strategy must be selected to enhance the performance of the system for day-to-day operation.
  • the mapping strategy pertains to an adjustment of parameters corresponding to one or more independent channels of a multi-channel cochlear implant or other hearing-enhancement system.
  • T-level electrical threshold
  • C-level comfort or "max" level
  • a speech processor then is programmed or "mapped" using one of several encoding strategies so that the electrical current delivered to the implant will be within this measured dynamic range; that is between the T- and C-levels.
  • T- and C-levels are established and the mapping is created, the microphone is activated so that the patient is able to hear speech and other sounds. From that point onwards, the tuning process continues as a traditional hearing test. Hearing-enhancement device users are asked to listen to tones of varying frequencies and amplitudes. The gain of each channel can be further altered within the established threshold ranges such that the patient is able to hear various tones of varying amplitudes and frequencies reasonably well.
  • fitting and tuning a hearing-enhancement system of any type so as to meet the needs of a particular patient is typically quite costly and very time consuming, both from the perspective of the hearing-impaired patient and the audiologist.
  • the functions of such a system are regulated by a large number of parameters, values for each of which typically must be determined so as to tune the system to provide optimal performance for the particular patient. In order to do so, the patient typically must be thoroughly tested with respect to each of set of parameter values. The number of tests generally increases exponentially as the number of system parameters increases. Additionally, the testing environment itself can be a factor when testing a user and setting system parameters.
  • perceptual devices such as cochlear implants or hearing aids, may be tuned to improve a user's hearing across a broad range of frequencies and/or noise volume levels. Testing in noise effectively and quickly exposes a person's hearing weakness, thus reducing overall test time and providing a more accurate model of their hearing weakness.
  • One aspect of the invention is to provide system and methods that utilize techniques for identifying the raw perceptual weaknesses of a subject in a considerably shortened time frame as compared to conventional techniques.
  • the invention relates to a computer-implemented method for generating a test set having test stimuli for testing a subject using a computer system having logic-based processing circuitry, the method including: selecting one or more features from among a plurality of features, wherein a measurable effect on the subject for each feature selected exceeds a predetermined threshold; based on the selected features, generating one or more sample classes; selecting test stimuli from each sample class for presenting to the subject, wherein the selecting step includes: initially choosing each stimulus from each sample class to present to the subject, wherein the chosen stimuli include the test stimuli; and for each sample class that has more than one stimulus, subsequently deselecting at least one stimulus from the - A -
  • test stimuli In an embodiment of the above aspect, at least one of the deselected test stimuli is deselected randomly. In another embodiment, the method further includes deselecting at least one stimulus associated with a non-selected feature to present to the subject. In still another embodiment, deselecting at least one stimulus associated with a non-selected feature is repeated iteratively. In yet another embodiment, iteratively deselecting at least one stimulus associated with a non-selected feature includes deselecting one less stimulus from a class including more than one stimulus.
  • iteratively deselecting at least one stimulus associated with a non-selected feature includes randomly deselecting a stimulus.
  • the method further includes initially presenting each of the test stimuli prior to selecting from among the plurality of features associated with each of the test stimuli.
  • selecting from among the plurality of features includes computing for a feature a product equal to a value assigned to the feature times a number of stimuli influenced by the feature.
  • the method further includes determining a measure of significance of each sample class on assessing the hearing capability of the patient. In another embodiment, the measure of significance of the i th class is equal to
  • n t is the number of stimuli contained in the i ⁇ th class
  • py is an empirically determined value based upon an assessment of patient responses to administered test stimuli.
  • the method further includes determining the number of times to present different test stimuli to the patient based upon a computed measure of patient hearing weakness, the measure, su being defined as
  • the method further includes presenting the test stimuli to the subject.
  • the presenting step further includes: selecting at least two test stimuli; transmitting a noise stimulus to the subject; transmitting the two test stimuli to the subject; and receiving at least one associated user response to the two test stimuli, wherein the associated user response is based at least in part on the test stimuli.
  • the method includes determining a relative weakness of a feature set for the subject, wherein the relative weakness is based at least in part on the noise stimulus and the two test stimuli.
  • the method includes adjusting a volume of the noise stimuli from the first volume to a second volume; and repeating the noise transmitting step, the test stimuli transmitting step, and the receiving step.
  • the adjusting step is based at least in part on the associated user response.
  • the first volume is greater than the second volume.
  • the associated user response is a non-response.
  • the at least two test stimuli are a phoneme pair.
  • the noise stimulus has a temporal structure and the test stimuli have a temporal structure, and the noise temporal structure substantially matches the test temporal structure.
  • the transmitting step and the receiving step together are a sequential testing step, and the sequential testing step includes transmitting a first test stimulus, then receiving a first associated user response, then transmitting a second test stimulus, then receiving a second associated user response.
  • the method includes determining a threshold noise volume. In another embodiment, the threshold noise volume is the first volume.
  • the invention relates to a computer-implemented method of testing a hearing-impaired patient using a computer system including logic-based processing circuitry, the method including the steps of: audibly presenting a plurality of phonemes to the patient, wherein each phoneme is selected from one of a plurality of phoneme sets corresponding to a predetermined feature selected for testing a hearing capability of the patient, wherein the plurality of phonemes are test stimuli; based on a response of the patient to each audibly presented phoneme, generating a first assessment of the hearing capability of the patient; and audibly presenting a second plurality of phonemes to the patient and generating a second assessment of the hearing capability of the patient based on patient response, wherein the second plurality of phonemes is selected by deselecting from the test stimuli at least one phoneme from each phoneme set that contains more than one phoneme, wherein at least one of the deselected phonemes is deselected randomly.
  • the method further includes presenting a third plurality of phonemes, wherein the third plurality of phonemes is selected by deselecting from the test stimuli at least two phonemes from each phoneme set that contains more than one phoneme, wherein at least one of the deselected phonemes is deselected randomly.
  • the method includes adjusting at least one operational parameter of a hearing-enhancement device based at least in part on a hearing performance of the subject with respect to different parameter values.
  • the invention in another aspect, relates to a computer-based system for generating a test set of test stimuli for testing a subject, the system including: at least one processor; an electronic memory having stored therein electronic data representing a plurality of features for testing a particular capability of the subject; a feature-selecting module configured to execute on the at least one processor for selecting one or more features from among the plurality of features, wherein a measurable effect on the subject of each feature selected exceeds a predetermined threshold; and a stimulus-selecting module configured to execute on the at least one processor for generating one or more classes of stimuli based upon the selected feature, and selecting at least one stimulus from each class for presenting to the subject; wherein the stimulus-selecting module is further configured to initially choose each stimulus from each class to present to the subject wherein the chosen stimuli include the test stimuli, and subsequently deselecting from the test set at least one stimulus from each class that includes more than one stimulus, wherein at least one of the deselected test stimuli is deselected
  • the stimulus- selecting module is further configured to deselect at least one stimulus associated with a non-selected feature to present to the subject.
  • stimulus-selecting module iteratively deselects at least one stimulus associated with a non-selected feature.
  • iteratively deselecting at least one stimulus associated with a non-selected feature includes deselecting one less stimulus from a class having more than one stimulus.
  • iteratively deselecting at least one stimulus associated with a non-selected feature includes randomly deselecting a stimulus.
  • the system is configured to initially present each of the predetermined number of stimuli prior to selecting from among the plurality of features associated with each of the predetermined number of stimuli.
  • the feature- selecting module is configured to select from among the plurality of features by computing for a feature a product equal to a value assigned to the feature times a number of stimuli influenced by the feature.
  • the system includes a significance-determining module for determining a measure of significance of each class of stimuli on assessing the hearing capability of the patient. In still another embodiment the measure of significance is equal to
  • the system includes a stimulus-presentment module for determining a number of times to present different stimuli to the patient based upon a computed measure of patient hearing weakness, the measure, Si, defined as
  • the invention relates to a computer-based system of testing a hearing- impaired patient with a test set, the system including: an audio unit for audibly presenting a plurality of phonemes to the patient, wherein each phoneme is selected from one of a plurality of phoneme sets corresponding to a predetermined feature selected for testing a hearing capability of the patient, wherein the plurality of phonemes include the test set; and a testing unit including at least one processor for generating a first assessment of the hearing capability of the patient based upon a response of the patient to each audibly presented phoneme; wherein the system audibly presents a second plurality of phonemes to the patient and generates a second assessment of the hearing capability of the patient based upon patient response, wherein the second pluralit
  • the invention in another aspect, relates to a computer-readable medium having embedded therein computer-executable code that, when downloaded and executed by a computer system, causes the computer system to: select one or more features from among a plurality of features, wherein a measurable effect on the subject of each feature selected exceeds a predetermined threshold; based on the selected features, generate one or more classes of stimuli; and select test stimuli from each class for presenting to the subject, wherein the selection of test stimuli includes initially choosing each stimulus from each class to present to the subject, and subsequently deselecting from the test stimuli at least one stimulus from each class that includes more than one stimulus, wherein at least one of the deselected phonemes is deselected randomly.
  • the invention in another aspect, relates to a computer-implemented method for generating a test set for testing a subject using a computer system including logic-based processing circuitry, the method including: selecting one or more features from among a plurality of features, the selecting including selecting one or more feature having a measurable effect on the subject, as determined based on a predetermined threshold; based on the selected features, generating one or more classes of stimuli; and selecting stimuli from one or more of said classes for presenting to the subject.
  • the invention in another aspect, relates to a computer-implemented method for generating a test set for testing a subject using a computer system having logic-based processing circuitry, the method including: selecting one or more features from among a plurality of features, wherein a measurable effect on the subject for each feature selected exceeds a predetermined threshold; based on the selected features, generating one or more classes of stimuli; and selecting stimuli from one or more of said classes for presenting to the subject, wherein selecting includes choosing a stimulus from a class to present to the subject wherein the chosen stimuli includes the test set, and subsequently deselecting at least one stimulus from the test set from each class that includes more than one stimulus.
  • the invention in another aspect relates to a method of assessing hearing characteristics of a subject, the method including: determining a hearing capability of the subject based on responses of the subject to a series of sounds presented to the subject, each sound corresponding to a presence, absence or irrelevance of a predetermined plurality of features; and assigning the subject to one of a predetermined plurality of classes based upon the responses of the subject, each of the plurality of classes being derived from hearing tests performed on a plurality of other subjects.
  • the method includes setting one or more parameter values of a hearing-enhancement device based on the class to which the subject is assigned.
  • the step of determining hearing capability includes identifying one or more of the plurality of features as contributing more than other of the plurality of features to a failure of the subject to correctly respond to the presentment of one or more, of the series of sounds.
  • a failure to correctly respond to a particular one of the series of sounds defines a feature error with respect to the one or more features corresponding to that particular one of the series of sounds, and further including generating a performance measure for the subject based upon a computed mean of feature errors.
  • the computed mean of feature errors equals a weighted mean, and further including computing the weighted mean, ⁇ , to be equal to
  • the step of assigning the subject to one of the predetermined plurality of classes includes computing a weighted contribution of each feature, the weighted contribution of a feature quantitatively measuring the contribution that the feature makes to the computed mean of feature errors.
  • computing the weighted contribution of a feature includes computing a value equal to
  • the invention relates to a system for tuning a hearing- enhancement device, the system including: a subject interface for rendering a series of sounds to a subject and for receiving from the subject a response to each of the sounds rendered, each sound corresponding to one or more features belonging to a predetermined plurality of features; and a processing unit communicatively linked, to said subject interface, the processing unit having: a hearing-capability module for determining a hearing capability of the subject based on the received responses of the subject to the series of sounds rendered; a class- assigning module for assigning the subject to one of a predetermined plurality of classes based upon the received responses, each of the plurality of classes being derived from hearing tests performed on a plurality of other subjects; and a tuning module for setting one or more parameters of a hearing-enhancement device based on the class to which the subject is assigned.
  • the hearing-capability module is configured to determine hearing capability of the subject by identifying one or more of the plurality of features as contributing more than other of the plurality of features to a failure of the subject to correctly respond to the presentment of one or more of the series of sounds.
  • a failure to correctly respond to a particular one of the series of sounds defines a feature error with respect to the one or more features corresponding to that particular one: of the series of sounds, arid wherein the hearing-capability module is further configured to measure a hearing performance of the subject, based on a computed mean of feature errors.
  • the computed mean of feature errors equals a weighted mean, and wherein the hearing-capability module is configured to compute the weighted mean, ⁇ , to be equal to
  • the class-assigning module is configured to assign the subject to one of the predetermined plurality of classes by computing a weighted contribution of each feature, the weighted contribution of a feature quantitatively measuring the contribution that the feature makes to the computed mean of feature errors.
  • the class-assigning module computes the weighted contribution of a feature by determining a value equal to
  • the invention relates to a computer-readable storage medium in which computer-readable code is embedded, the computer-readable code configured to cause a computing system to perform the following steps when loaded on and executed by the computing system: determine a hearing capability of a subject based on responses; of the. subject to a series of sounds presented to the subject, each sound corresponding to a presence, absence or irrelevance of a predetermined plurality of features; and assign the subject to one of a predetermined plurality of classes based on his responses, each of the plurality of classes being derived from hearing tests performed on a plurality of other subjects.
  • the storage medium includes: computer- readable code for causing the computing system to set one; or more parameter values of a hearing-enhancement device based upon the class to which the subject is assigned.
  • the step of determining hearing capability includes identifying one or more of the plurality of features as contributing more than other of the plurality of features to a failure of the subject to correctly respond to the presentment of one or more of the series of sounds.
  • a failure to correctly respond to a particular one of the series of sounds identifies a feature error with respect to the one or more features corresponding to that particular one of the series of sounds, and further includes measuring a hearing performance of the subject based on a computed mean of feature errors.
  • the computed mean of feature errors equals a weighted mean, and further including computing the weighted mean, ⁇ , to be equal to
  • W 1 is a weight assigned to the i th feature of the plurality of features and n t is the number of feature errors with respect to the i th feature.
  • the step of assigning the subject to one of the predetermined plurality of classes includes computing a weighted contribution of each feature, the weighted contribution of a feature quantitatively measuring the contribution that the feature makes to the computed mean of feature errors.
  • the computer-readable storage medium of claim 59 wherein computing the weighted contribution of a feature includes computing a value equal to
  • Contribution(fi) is the weighted contribution of the i th feature.
  • the invention includes an article of manufacture having a computer-readable medium with computer-readable instructions embodied thereon for performing the methods described in the preceding paragraphs.
  • a computer-readable medium such as, but not limited to, a floppy disk, a hard disk, an optical disk, a magnetic tape, a PROM, an EPROM, CD-ROM, DVD-ROM or downloaded from a server.
  • the functionality of the techniques may be embedded on the computer-readable medium in any number of computer- readable instructions, or languages such as, for example, FORTRAN, PASCAL, C, C++, Java, PERL, LISP, JavaScript, C#, TcI, BASIC and assembly language.
  • the computer- readable instructions may, for example, be written in a script, macro, or functionally embedded in commercially available software (such as EXCEL or VISUAL BASIC).
  • FIG. 1 is a schematic diagram of a system for generating a test set for testing a subject, according to one embodiment of the invention.
  • FIG. 2 is a schematic diagram of a system of testing a hearing-impaired patient, according to another embodiment of the invention.
  • FIG. 3 is a flowchart of exemplary steps in a method for generating a test set for testing a subject, according to yet another embodiment of the invention.
  • FIG. 4 is a flowchart of exemplary steps in a method for testing a subject utilizing noise stimuli, according to another embodiment of the invention.
  • FIG. 5 is a schematic diagram of an environment in which a system for tuning a hearing-enhancement device, according to one embodiment of the invention, can be utilized.
  • FIG. 6 is a schematic diagram of another environment in which a system for tuning a hearing-enhancement device, according to a different embodiment of the invention, can be utilized.
  • FIG. 7 is a more detailed schematic view of a system for tuning a hearing-enhancement device, according to one embodiment of the invention.
  • FIG. 8 is a flowchart of exemplary steps in a method of testing a hearing-impaired subject, according to still another embodiment of the invention.
  • the system 100 illustratively includes one or more processors 102 comprising registers, logic gates, and other data processing circuitry (not explicitly shown) for executing processor-executable instructions.
  • the system 100 also illustratively includes an electronic memory 104 for electronically storing processor-executable instructions and data.
  • the system 100 illustratively includes a feature-selecting module 106, the operative features of which are described more particularly below and which is communicatively linked to the one or more processors 102.
  • the system 100 further illustratively includes a stimulus-selecting module 108, the operative features of which are also described more particularly below.
  • the stimulus- selecting module 108 is communicatively linked to the one or more processors 102.
  • each of the feature- selecting module 106 and the stimulus- selecting module 108 can be implemented in a combination of the processing circuitry of the one or more processors and processor-executable instructions that, when loaded to and executed by, one or more processors performs the operations, procedures, and functions described herein.
  • one or both the feature-selecting module 106 and the stimulus-selecting module 108 can be implemented in dedicated hardwired circuitry configured to perform the same operations, procedures and functions.
  • processing circuitry, hardware and/or firmware may be utilized to implement the feature-selecting module 106 and the stimulus- selecting module 108.
  • the system 100 stores in the memory 104 electronic data representing a plurality of features for testing a particular capability of the subject.
  • the feature- selecting module 106 executes on the one or more processors and selects one or more of the features from among the plurality of features.
  • the feature-selecting module 106 is configured to select those features that have a significant effect on the subject; that is, each feature selected exceeds a predetermined threshold.
  • the stimulus- selecting module 108 executing on the one or more processors generates one or more classes of stimuli based on the selected feature.
  • a stimulus is any signal, question, or other other-response eliciting element that is conveyed to a subject to elicit one or more responses that can be used to determine a capability, condition, or attribute of the subject.
  • Stimuli can be grouped into distinct classes corresponding to one or more features, wherein the features are those attributes associated with certain stimuli. As described herein, the subject's ability to discern certain features when presented with a corresponding stimulus indicates a particular capability or characteristic of the subject.
  • the stimulus- selecting module 108 selects at least one stimulus from each class for presenting to the subject.
  • the stimulus-selecting module 108 is further configured to initially choose each stimulus from each class to present to the subject, and to subsequently select at least one less from each class that comprises more than one stimulus.
  • at least one of the stimuli subsequently selected from each class that comprises more than one stimulus is selected randomly by the stimulus-selecting module 108.
  • a stimulus-selecting module may be presented with fewer than all stimuli if the tester determines a priori that certain stimuli are of lesser significance.
  • Stimuli can be used to test particular capabilities of a subject, such as a patient. Specific stimuli can be associated with particular features that characterize the particular capabilities. Thus, generally, the purpose of eliciting a response of the subject to the stimuli is to identify features with respect to which the subject is weak. For example, in the context of testing hearing, two distinctive feature sets have been proposed. The first is based on the articulatory positions underlying the production of speech sounds. The other is based on the acoustic properties of various speech sounds.
  • the fundamental source features can be further characterized on the basis of whether the speech sounds are vocalic or non-vocalic.
  • Vocalic speech corresponds to speech sounds associated with vowels. Accordingly, such speech sounds correspond to a single periodic source, the onset of the speech not being abrupt; otherwise the speech sound can be characterized as non-vocalic.
  • the fundamental source features also can be characterized on the basis of whether the speech sounds are consonantal or non-consonantal.
  • Consonantal speech sounds correspond to sounds associated with consonants. Such speech sounds are characterized by the presence of zeros in the associated spectrum of the sounds.
  • the secondary consonantal source features can be further characterized on the basis of whether the speech sounds are interrupted or continuant. Continuant speech sounds, are also characterized as semi- vowels, because of their similar sound quality. There is little or no friction with continuant speech sounds as the air passes out freely through the mouth of the speaker. A continuant speech sound is produced with an incomplete closure of the vocal tract. Interrupted speech sounds, by contrast, end abruptly.
  • the secondary consonantal features can also be characterized on the basis of whether the speech sounds are checked or unchecked. Checked speech sounds, typified by some Far Eastern and African languages, are characterized by abrupt termination as opposed to gradual decay, whereas unchecked speech sounds are characterized by gradual decay.
  • secondary consonantal features can be characterized as strident or mellow.
  • the former typically has an irregular waveform, whereas the latter typically has a smooth waveform.
  • a secondary consonantal feature characterized as mellow also has a wider autocorrelation function relative to a corresponding normalized strident feature. Secondary consonantal features can also be classified according to whether the sound is voiced or voiceless.
  • the resonance features can be further characterized on the basis of whether the speech sound is compact or diffuse.
  • a compact feature is associated with sound having a relative predominance of one centrally located format region, whereas a diffuse features implies sound having one or more non-central formats.
  • the resonance features can also be characterized as grave or acute. Speech sounds that are characterized as grave are low- frequency dominant low frequency, whereas those characterized as acute are high-frequency dominant. Additionally, resonance features can be characterized as flat or plain, depending on whether the there is a downward shift of some or all formats, typically associated with vowels and a reduction in lip orifice of the speaker. [0035]
  • the resonance features also can be further characterized as sharp or plain, the latter characterizing speech sounds whose second and/or higher formats rise.
  • resonance features can also be characterized as tense or lax, depending on the amount and duration of the energy of the sound.
  • the resonance features also can be classified according to whether the speech sound is characterized as having a nasal format or a nasal murmur.
  • the distinctive speech features and their potential acoustic correlates are further described in R. Jakobson, G. M. Fant, and M. Halle, PRELIMINARIES TO SPEECH ANALYSIS: THE DISTINCTIVE FEATURES AND THEIR CORRELATES (MIT Press, Cambridge; 1963), which is incorporated herein by reference in its entirety.
  • a subject weak with respect to a certain class of stimuli characterized for example by an inability to recognize or respond to the stimuli, is typically weak with respect to one or more of the features associated with that class of stimuli.
  • the size of test set of stimuli is combinatorial with respect to the number of features. Accordingly, it is advantageous to identify the features that play a significant role in determining the subject's hearing ability.
  • the stimuli can be classified into a smaller number of classes based on the significant features. The number of classes of stimuli increases exponentially with the number of features; if the less significant features are not eliminated, many of the classes will be empty.
  • the system 100 is configured to identify the vital features so as to reduce the number of stimuli classes and allow each class to contain a meaningful set of stimuli. Also, by emphasizing on the vital features, it is expected that the majority of a subject's weaknesses can be ascertained in a shorter period of time than is conventionally the case. Since the brain of a human being is generally adept at recognizing a stimulus from only partial information, it will be sufficient to rectify the weaknesses caused by the more influential features in, for example, the context of tuning a hearing-enhancement device.
  • the feature-selecting module 106 is configured to determine whether the effect of a feature on a subject exceeds a predetermined threshold according to the following: fl if v Xm > ⁇ [ ⁇ otherwise where ⁇ is the predetermined threshold, V 7 - is the number of different values assumed by the ith feature for representing the effect of the test stimuli (e.g., +1 if the feature is present in the stimulus, -1 if it is absent, and 0 if the feature is irrelevant), m ; is the number of different stimuli influenced by the Mh feature. Accordingly,/ is a quantitative measure of the role of the Mi feature. During testing in a preferred embodiment a threshold value of 7 produced excellent results.
  • threshold values may also be used to produce relevant phoneme discrimination.
  • the feature selecting module can be configured differently such that the effect of the features can be less than a predetermined threshold.
  • the stimulus- selecting module 106 can be configured to determine the significance of the i th class of stimuli. In a test, the number of different stimuli chosen from each class is typically proportional to the significance of that class. More particularly, the stimulus- selecting module 106 can be configured to compute the following measure of significance of a class of stimuli:
  • n is the number of different stimuli in the i th class
  • py be the significance of the/' 1 stimuli in the i th class on the stimuli-recognition ability of the subject. Therefore, significance of the i th class of stimuli on the recognition ability of a patient is given by the above calculation as performed by the stimulus selecting module 106.
  • the stimuli selecting module can be configured differently to compute a measure of significance.
  • the generation of a test set (i.e., the selected stimuli), as determined by the system 100 according to one embodiment, then depends on three factors: first, the significance of a class of stimuli, as discussed above; second, the weakness of the subject with respect to particular features (in many settings, the weakness becomes apparent within a few tests, and thus the subsequent presentation of stimuli becomes dependent on the weaknesses); and third, the mitigation of testing error, which according to the present invention is achieved by randomized stimuli selection, as described more particularly below.
  • the stimulus- selecting module 108 can be configured to quantitatively assess the weakness with respect to one or more features. Specifically, the stimulus-selecting module 108 can be configured to compute the following:
  • the stimuli selecting module can be configured differently to assess the weakness with respect to one or more features.
  • the stimulus selecting module is utilized to reduce the number of classes and the number of stimuli.
  • One of the ordinary skills in the art is to recognize that the number of stimuli could be reduced a priori based on perceived lower stimuli significance. However, this may result in compromise of the testing quality.
  • the operative features of the system 100 are now described in the specific context of generating a test set or selected stimuli for testing the hearing capabilities of a patient.
  • the test set can be utilized in fitting or "tuning" hearing-enhancement devices, such as hearing aids and cochlear implants, mobile telephones or wireless/Bluetooth devices, etc.
  • hearing-enhancement devices such as hearing aids and cochlear implants, mobile telephones or wireless/Bluetooth devices, etc.
  • the functions of such a device are regulated by a large number of parameters, values for each of which must be determined so as to tune the device to provide optimal performance for a particular patient.
  • Each parameter is assigned one of many values.
  • Conventional techniques for determining the optimal set of parameter values for each patient is difficult and time consuming. A patient typically must be thoroughly tested in order to ascertain the parameter values that yield optimum device performance for the patient.
  • the goal of testing the hearing-impaired patient is to ascertain his raw hearing ability independent of context and background knowledge.
  • a series of consonant phonemes can be presented, as stimuli, and the patient's response assessed so as to identify any weakness in his hearing.
  • Different parameters of the hearing-enhancement device can be adjusted accordingly.
  • the operative features of the system 100 permit the patient to be tested according to an adaptive method, according to a particular embodiment.
  • audible renderings of phonemes are presented to the patient, rather than words or phrases.
  • the phonemes are selected from a set of fourteen consonant phonemes.
  • consonant phonemes are preferably used rather than vowel phonemes, because the latter are typically too easily perceived by a patient and do not reveal sufficient information pertaining to the patient's hearing capability.
  • the system enables the number of phonemes utilized in testing to be reduced, affording a significant savings in time and resources without compromising the quality of the testing performed.
  • the patient's strengths and weaknesses are assessed based on the patient's response to phonemes corresponding to the different features represented by each.
  • a phoneme is characterized by the presence, absence or irrelevance of a set of nine features: Vocalic, Consonantal, Compact, Grave, Flat, Nasal, Tense, Continuant, and Strident.
  • the feature- selecting module 106 can be configured to operate on features arranged hierarchically. Those features higher in the hierarchy are those that potentially have a greater effect because the failure to perceive these features affects the perception of a greater number of phonemes. Thus, those features ranked higher in this hierarchy provide a more comprehensive measure of hearing loss as compared to words or phrases.
  • Each phoneme can be associated with a corresponding percentage of proportional occurrence of the phoneme in the English language. (See, e.g., L. Shriberg and R. Kent, Linical Phonetics, Boston: Allyn & Bacon (2003), incorporated here in its entirety.)
  • the presence, absence, and irrelevance of a feature with respect to each phoneme can be denoted, respectively, as 1, -1, and 0.
  • the fourteen consonant phonemes used in testing and the corresponding constituent features of each are: neme s Vocalic Cons. Compact Grave Flat Nasal Tense Cont. Strident n -1 1 -1 -1 0 1 0 0 0 t -1 1 -1 -1 0 -1 1 -1 0
  • the features Vocalic and Consonantal remain the same with respect to all fourteen phonemes.
  • the features Tense, Continuant, and Strident do not make a substantial difference to hearing ability, as has been verified empirically.
  • the feature Flat does not influence any of the fourteen phonemes.
  • the following exemplary pseudo-code illustrates a procedure that can be implemented by the feature-selecting module 108 for creating a hierarchy of the nine features: This hierarchy is derived in R. Jakobson, G. Fant, and M. Halle, Preliminaries to Speech Analysis, Cambridge, MA: The MIT Press, 1963, which is incorporated herein in its entirety.
  • the consonantal sound " ⁇ " referred to as the "voiced dental non-sibilant fricative" in the International Phonetic Alphabet, is utilized in the line reading 'If Mellow -> "/ ⁇ /" as in THat.'
  • v(/th feature) the number of different values assumed by the i th feature for representing the test stimuli.
  • the m(i th feature) the number of different stimuli influenced by the i th feature.
  • Nasal can affect at most 4 phonemes; that is, perceiving Nasal incorrectly can affect at most 4 phonemes.
  • the m(i th feature) is dependent on the feature hierarchy, not just on the number of +1, -1 or 0 values assumed for the different phonemes.
  • the feature hierarchy is a tree-like structure with the features occurring at the non-leaf nodes and the phonemes occurring at the leaf nodes. To reach a phoneme at a leaf node, one has to traverse the path (or branches) from the tree top (or root) down to the phoneme. One phoneme can be reached via exactly one path, so if a mistake occurs somewhere in the middle of the path (e.g., +1 is chosen for Nasal instead of-1), then it is impossible to reach that phoneme. More significant features, such as Vocalic and Consonantal, are higher up in the tree-like hierarchy, while the less significant features, such as Continuant and Strident, are at the lower levels.
  • m(Strident) is much less than m(Vocalic).
  • m(i th feature) calculates the size of the largest number of phonemes on one side of the tree starting from the non-leaf node corresponding to the i th feature.
  • the phonemes are classified based on the values assumed by these three features.
  • Each of these features can assume three values:
  • each unique combination of values of the selected features forms a class. So, the combination -1, 1, 1 is a distinct class and so is the combination -1, 1,-1. Most of these classes do not contain any phoneme. Only five of them do. For example, the class 1, 1, 1 does not contain any phoneme, the class -1, 1, .1 contains only 'm', while the class -1, 1, -1 contains ⁇ b, P, v, f ⁇ .
  • the remaining four phonemes ⁇ k, g, sh, j ⁇ have three important features as 1, 0, -1. Though that does not belong to any of the eight combinations sought, one of them can be used to test for ⁇ 1 1 -l> or ⁇ l -1 -1>, especially given that there is no phoneme that has any of these two combinations.
  • a minimal set of consonant phonemes contains five phonemes - m, n, and one each from the subsets ⁇ b, p, v, f ⁇ , ⁇ t, s, d, z ⁇ , and ⁇ k, g, sh, j ⁇ . The set is devoid of redundancy with respect to the features of interest.
  • the resulting test set is suitable for testing under ideal conditions. In practice, however, testing errors do occur. Thus, using the system 100 in the context of testing hearing, the hearing-impaired patient is tested with a few redundant phonemes in order to compensate for testing errors. Moreover, a patient may have hearing loss only in particular frequency ranges, which may not be evident from feature analysis. Thus, the testing uses redundant phonemes to mitigate these problems.
  • testing under less than ideal conditions may also reduce test time and more quickly identify a patient's weakness in ability to distinguish phonemes. This may include testing the resulting test set of phonemes in varying levels of different kinds of noise.
  • One proposed testing method includes testing with a prescribed set of phonemes in varying levels of different kinds of noise.
  • Types of noise that may be utilized include white noise, pink noise, other broadband noises generated by filtering white noise and/or shaping the temporal structure of the noise envelop, environmental noises (sirens, alarms, machine noise, etc.), unintelligible/indistinguishable speech ("speech babble,” as commonly encountered in a room containing a large number of people), as well as distorted speech such as that obtained through frequency and/or temporal compression, translation or any other manipulation, or by altering the phase of different components.
  • the threshold level is first determined based on a level of phoneme or feature errors for a particular user.
  • test stimuli such as a series or pairs of phonemes (e.g., 'b' and V, 'p' and 'f, 'm' and 'n', etc.) may be transmitted to a user, while the user is also exposed to the noise stimulus.
  • the series or pairs may include phonemes having similar features, as well as unrelated features.
  • the user is then asked to distinguish between the series or pairs of phonemes by selecting a proposed response, typing a response, speaking a response, etc.
  • the background noise level may be varied until the patient fails to distinguish between the test stimuli.
  • the amount of noise required to make the patient fail to distinguish the test stimuli provides an indication of the nature of weakness in the hearing ability of the patient.
  • the speech perception task may be made more difficult by other factors, such as by systematically reducing the gain (output level), by adding additional tasks that require the listener's attention, etc.
  • testing may be performed with each phoneme.
  • only certain pairs of phonemes may be utilized, thereby reducing the testing time.
  • the noise stimulus is gradually reduced from the threshold level, and the phonemes are not previously identified to the patient. This prevents the patient from knowing what the testing entails, which may provide a more accurate test result.
  • the noise stimulus begins at the threshold level, the patient is unable to know the phonemes until he can actually perceive them.
  • testing with the noise stimulus beginning at the threshold level may be desirable, testing with a lower noise stimulus level may also be utilized.
  • Various strategies for varying the noise may be employed so that the actual time for the entire test is less than the traditional methods of testing with the prescribed set of phonemes. These strategies include, but are not limited to, isolation of a break point of a single feature.
  • Testing systems may employ one type or several types of noise, as described briefly above.
  • the nature of the noise may have a significant impact on the end results - for example, a phoneme pair intended to test perception of signals at high frequencies may necessitate the use of a noise stimulus with sufficient energy in the same frequency region. In contrast, other phoneme pairs may be tested more accurately with a noise having greater temporal variation.
  • the noise stimulus has a temporal structure that substantially matches the temporal structure of the test stimuli.
  • the type of responses expected may also impact the accuracy of the testing. While it may be less complex to utilize a closed set of responses from which a user may choose, users may be able to guess the correct response, providing less accurate results.
  • the phonemes are selected based on the patient's weaknesses. In the first test, all fourteen phonemes are presented, since initially there is no assessment of the patient's weaknesses. In the next two tests, eleven phonemes (three from each of the three sets described above) are presented. In the fourth test onwards, eight phonemes (two from each of the three sets) are presented. In the second and third tests, two phonemes from each set are chosen based on the patient's performance model, which captures the feature errors in the previous test. The remaining one phoneme from each set is chosen randomly from the remaining two phonemes. From the fourth test onwards, one phoneme from each set is chosen based on the patient's performance model, while the other phoneme is chosen randomly from the remaining three phonemes. The strategy is summarized as follows:
  • FIG. 2 is a schematic diagram of a computer-based system 200 for testing a hearing-impaired patient, according to another embodiment of the invention.
  • the system 200 illustratively includes an audio unit 202, for audibly presenting a plurality of phonemes to the patient, wherein each phoneme is selected from one of a plurality of phoneme sets corresponding to a predetermined feature selected for testing a hearing capability of the patient. Additionally, if testing in noise is desired, the audio unit may present the selected noise stimuli.
  • the system 200 further includes a testing unit 206 comprising at least one processor 208 for executing the procedures described above in generating assessments of the hearing capability of the patient based upon patient responses audibly-presented phonemes.
  • the audio unit 202 can comprise, for example, a speaker, headphones, or other electromechanical transducer (not explicitly shown) for generating sound signals that can be played or otherwise rendered to the patient.
  • the audio unit 202 can be a hearing-enhancement device, a telephone, wireless phone, cellular phone, or the like.
  • the audio unit 202 can optionally include a microphone or other acoustical transducer for converting audible responses of the patient into electrical signals that are conveyed to the testing unit 206.
  • the system 200 can optionally include a separate patient- response device 204, such as a hand-held push-button device, a keypad or the like that can be used by the patient in response to audibly-presented phonemes. The purpose of these different arrangements is to permit the system 200 to present the plurality of phonemes to which the patient responds so as to assess hearing capabilities of the patient.
  • the system 200 can also include a recorder 210 for recording the audible responses or signals conveyed to the testing unit 206 by the patient using the optional patient-response device 204.
  • the testing unit 206 generates a first assessment of the hearing capability of the patient based upon a response of the patient to each audibly presented phoneme, according to the procedures described above. Additionally, the testing unit 206 may generate a predetermined noise stimuli if testing with background noise is desired.
  • the system 200 subsequently, audibly presents a second plurality of phonemes to the patient and generates a second assessment of the hearing capability of the patient based upon patient response.
  • the second plurality of phonemes is selected by the testing unit 206 choosing at least one less phoneme from each phoneme set that contains more than one phoneme, one of the phonemes selected from each phoneme set containing more than one phoneme being selected randomly by the testing unit.
  • background noise may be utilized during the entire test, or any portion thereof.
  • the system 200 optionally can include one or more databases 212 for storing the plurality of phonemes and noise stimuli.
  • the audio unit 202 is shown as communicatively linked directly, wirelessly or through a wire-line connection, with the testing unit 206, it will be readily apparent to one skilled in the relevant art that the system can be communicatively linked to the audio unit through one or more intermediate communication nodes of voice-based network or data communications network, such as the Internet.
  • FIG. 3 illustrates certain method aspects of the invention.
  • FIG. 3 is a flowchart of exemplary steps in a method 300 of generating a test set for testing a subject using a computer system comprising logic-based processing circuitry.
  • the method 300 illustratively includes, after start at block 302, selecting one or more features from among a plurality of features at block 304.
  • the selecting more particularly comprises selecting one or more features having a measurable effect on the subject, as determined based on a predetermined threshold.
  • the method 300 further includes generating one or more classes of stimuli based on the selected features at block 306. Additionally, the method includes selecting stimuli from one or more of the classes at block 308a. In parallel to step 308a, if testing is to be performed with noise stimulus, the appropriate noise stimulus is selected 308b.
  • a noise stimulus that generally matches or approximates the temporal structure of the test stimuli may be desirable to accurately test the perceptual weakness of a subject.
  • a universal noise stimulus e.g., speech babble
  • the noise stimulus need not be selected in parallel with the test stimuli, as depicted in FIG. 3, but instead may be a predefined noise stimulus, regardless of test stimuli.
  • the selected test stimuli and noise stimulus are then presented to the subject at block 310.
  • the method illustratively concludes at block 312.
  • FIG. 4 depicts a procedure for testing a subject utilizing background noise 350.
  • the procedure begins by selecting at least two test stimuli 352, in accordance with the methods described above. Thereafter, the noise stimulus (selected in Step 304a of FIG. 3) is transmitted to the subject 354.
  • This noise stimulus may be transmitted or presented at a threshold volume, that is, the maximum volume that a subject may comfortable listen to. This threshold volume may be predetermined at set at the time of testing, or may be determined prior to presentation of the noise stimulus.
  • the test stimuli is transmitted to the subject 356 and the testing system waits for a response.
  • the response from the subject is received 358, either via a spoken sound, keyboard signals, or other forms described above.
  • the subject response may be a non-response or a delayed response, thus signaling the testing system that the subject may be fatigued and that testing should be terminated.
  • the method may enter a sequential testing loop 360, where a first stimulus is used to elicit a first response from the user, followed by a second stimulus, a second response, and so on.
  • the processor receiving the user responses stores the various test stimuli and associated responses for later determination of subject perceptual weakness.
  • the system processor determines the relative perceptual weakness of the subject 362, in accordance with the methods described herein. At this point, the testing may conclude and the information obtained therefrom may be used to optimize or otherwise tune a hearing or other perceptual device. Exemplary tuning procedures and systems include those described in U.S. Patent No. 7,206,416, the disclosure of which is hereby incorporated by reference herein in its entirety. Alternatively, the procedure may continue. In one embodiment, the volume of the noise stimuli 364 may be adjusted and the testing procedure repeated 366. In certain cases, the new, adjusted noise stimuli volume may be increased or decreased, relative to the previously- tested volume. Adjustment of the noise volume may aid in determining the strengths and weaknesses of a subject, thus reducing overall test time.
  • a method of generating a test set for testing a subject using a computer system comprising logic-based processing circuitry includes selecting one or more features from among a plurality of features, wherein a measurable effect on the subject of each feature selected exceeds a predetermined threshold. Additionally, method includes generating one or more classes of stimuli based on the selected features. The method further includes selecting stimuli from one or more of the classes for presenting to the subject, wherein selecting comprises choosing a stimulus from a class to present to the subject, and subsequently selecting at least one less from each class that comprises more than one stimulus.
  • Still another embodiment of a method for generating a test set for testing a subject using a computer system comprising logic-based processing circuitry includes selecting one or more features from among a plurality of features, wherein a measurable effect on the subject of each feature selected exceeds a predetermined threshold, and based on the selected features, generating one or more classes of stimuli.
  • the method further includes selecting stimuli from each class for presenting to the subject, wherein selecting comprises initially choosing each stimulus from each class to present to the subject, and subsequently selecting at least one less from each class that comprises more than one stimulus, at least one of the stimuli subsequently selected from each class that comprises more than one stimulus being selected randomly.
  • FIG. 5 is a schematic view of an exemplary environment 400 in which system 402 for tuning a hearing- enhancement device, according to one; embodiment, can be utilized.
  • the exemplary environment 400 illustratively comprises, in addition to the system 402, an audio unit 404 that delivers sound signals to the system.
  • the audio unit 404 can comprise, for example, a speaker, headphones, or other electromechanical transducer (not explicitly shown) for generating sound signals in response to electrical signals that can be conveyed from the system 402 to the audio unit, the sounds being rendered then to a user of the audio unit.
  • the audio unit 404 can be, the hearing-enhancement device that is to be tuned.
  • the hearing-enhancement device can be a separate device that also connects to the system 402.
  • the audio unit 404 can optionally include a microphone or other acoustical transducer for converting acoustic signals generated by the subject of the audio unit 404 into electrical signals that are conveyed to the system.
  • The, exemplary environment 400 might additionally or alternately include a separate subject-response unit 406, such as a computer terminal for. presenting a graphical of other subject interface with which a subject interacts using a keyboard and/or computer mouse (neither explicitly shown).
  • the system 402 includes a subject interface 108 configured to communicatively link the system to the audio unit 404.
  • the exemplary environment 400 additionally or alternatively includes separate subject-response unit 406, then the subject interface 408 (or a corresponding one) also can be configured to communicatively link the system 402 to the subject-response unit.
  • the subject interface 408 can be used in playing the series of sounds that are presented to the subject to a subject and for receiving, from the subject a response to each of the: sounds played.
  • system 402 further includes a processing unit 410 for effecting the operative processes, procedures, and functions that are described more particularly below.
  • the system 402 can also, include one or more databases; 412 for storing the plurality of features and/or sound signals, such as phonemes, words, etc., that correspond to the presence, absence or irrelevance of the plurality of features.
  • the system 402 is shown as communicatively linked directly, wirelessly or through a wire-line connection, with the audio unit 404, it will be readily apparent to one skilled in the relevant art that the system can be communicatively linked to the audio unit through one or more intermediate nodes.
  • FIG. 6 illustrates one such embodiment in which a system 502 for tuning a hearing- enhancement device is linked to an audio unit 504 through a data communications network 506, such as a local-area network (LAN), a wide-area network (WAN), or a plurality of connected network such as the Internet.
  • a data communications network 506 such as a local-area network (LAN), a wide-area network (WAN), or a plurality of connected network such as the Internet.
  • the system 502 and audio unit 504 can be communicatively linked through a public- switched telephony network.
  • a hearing-enhancement device being tuned can also serve as the audio unit that connects to the system, or alternatively, the hearing-enhancement device and audio unit can be separate devices.
  • the system 502 can optionally connect separately to the audio unit 504 and to a hearing-enhancement device that is tuned by the system through the same network 506 or even a separate one.
  • the processing unit 410 of the system 402 includes a hearing-capability module 414.
  • the system 402 further illustratively includes a class-assigning module 416 communicatively linked to the hearing-capability module 414.
  • the system illustratively includes a tuning module 418.
  • One or more of the hearing-capability module 414, class-assigning module 416, and a tuning module 418 can be implemented in a combination of logic-based circuitry and processor-executable code.
  • the processing unit 410 can be a general-purpose computer or application- specific computer having one or more processors implemented with registers, arithmetic-logic units, control units, and/or other logic- based circuitry.
  • the hearing-capability module 414, class- assigning module 416, and a tuning module 418 can be implemented in dedicated, hardwired circuitry configured to operate cooperatively with other elements of the system 402.
  • the subject interface 408 plays a series of sounds to the subject and, in response thereto, receives from the subject a response to each of the sounds played.
  • Each sound corresponds to the presence, absence or irrelevance of a predetermined plurality of features.
  • the hearing-capability module 414 of the processing unit 410 determines a hearing capability of the subject based on the received responses to the series of sounds played.
  • the class-assigning module 416 assigns the subject to one of a predetermined plurality of classes based upon the received responses. Each of the plurality of classes consists of none, one or more subjects based on their hearing characteristics as assessed from the hearing tests performed on them.
  • the tuning module 418 operates by setting one or more parameters of the hearing-enhancement device based on the device parameter settings of a similar subject in the class to which the subject is assigned.
  • the series of sounds presented by the system 402 to the subject are phonemes.
  • the system 402 can include a recording device (not shown) to record not only the subject's response but the response time of the subject. Based on the responses, the system 402 can incrementally build a model from which the strengths and weaknesses of the subject can be determined.
  • a phoneme is the smallest unit of speech that distinguishes meaning. Words and sentences are a combination of phonemes in a particular order.
  • the system 402 is configured to present to the subject phonemes selected from a set of fourteen consonant phonemes (as distinct from words or sentences), the phonemes being those identified in the Iowa Medial Consonant Recognition Test. Vowel phonemes need not be utilized by the system, primarily because such are considered too easy to perceive and thus much less likely to be useful indicators of the nature of hearing capabilities or hearing loss.
  • the system 402 provides an analysis of the subject's strengths and weaknesses in terms of a predetermined set of features associated with each phoneme. Phoneme characterization, percentage of proportional occurrence in the English language, and feature hierarchical arrangement are described above. As shown in U.S. Patent No. 7,206,416 to Krause, et al., these features provide a more comprehensive measure of the subject's nature of hearing loss as compared to words or sentences.
  • the system 402 is configured to classify subjects based on their performance in the three features: Compact, Grave, and Nasal.
  • An objective of the system 402 is to significantly reduce resource expenditures and time in testing.
  • One way to achieve this is by using knowledge obtained through testing of other subjects or subjects in the past.
  • the knowledge can be used to suggest parameter values for tuning, a hearing-enhancement device for a particular subject. Two subjects often have similar hearing characteristics, in which case it is highly probable that the optimum parameters values for one of the subjects are the same or very nearly so for the other.
  • Using knowledge obtained from previously-tested subjects can thus contribute significantly to the goal of improving the effectiveness and efficiency of testing a subject and setting parameter values by suggesting those values most likely to be the optimal settings. Utilizing the suggested parameter values can reduce testing and tuning times significantly.
  • a subject's performance in a test can be measured by the number of phonemes that the subject fails to perceive correctly. This, however, fails to capture the subject's strengths and weaknesses because many phonemes share similar features. For example, the phonemes T and 'p' differ only in one out of the nine features called Continuant. A person who fails to perceive 'p' due to an error in any feature other than Continuant will likely fail to perceive 'f as well. The converse is also true. Thus counting the number of phoneme errors is likely to be meaningless because feature errors give rise to phoneme errors. For the same reason, in order to reduce the phoneme errors, the system 402 is configured to concentrate instead on feature errors.
  • the hearing-capability module 414 is configured to determine hearing capability of the subject by identifying one or more of the plurality of features as contributing more than the others to a failure of the subject to correctly respond to the presentment of one or more of the series of sounds. More particularly, the hearing- capability module 414 can be configured to treat a failure to correctly respond to a particular one of the series of sounds as a feature error. Each feature error corresponds to a particular one of the series of sounds, which preferably are each phonemes.
  • the hearing-capability module 414 can be further configured to measure the hearing performance of the subject based on a computed mean of feature errors.
  • the computed mean of, feature errors can be a weighted mean
  • the hearing- capability module 414 can be configured to compute the weighted mean, ⁇ , to be equal to
  • W 1 is a weight assigned to the i th feature of the plurality of features and n t is the number of feature errors with respect to the i th feature.
  • An experimentally-determined set of weights for the nine above-listed features - Vocalic, Consonantal, Compact, Grave, Flat, Nasal, Tense, Continuant, and Strident - is ⁇ 0.151785714, 0.151785714, 0.098214286, 0.0, 0.142857143, 0.125, 0.125, 0.0625 ⁇ .
  • the system 402 can classify a subject based on the subject's strengths and weaknesses in perceiving phonemes.
  • the objective is to classify subjects with similar hearing characteristics in the same class and subjects with different hearing characteristics in different classes. Two subjects are considered to have similar hearing characteristics if the same feature contributes more errors than another feature.
  • only three features - Compact, Grave, and Nasal - contribute significantly in determining the nature of hearing loss for a majority of subjects.
  • the weights assigned to these features from the experimentally determined set are 0.142857143,0.098214286, and 0.142857143, respectively.
  • the class-assigning module 416 can be configured to rank the three features based on their weighted contribution to the total weighted error. Thus for the i th feature,/ ⁇ , among the plurality of features, the class-assigning module 416 can be configured to compute the weighted contribution to be
  • a subject belongs to one of the six classes. It is noted that when testing begins initially, the class to which the subject belongs is unknown. After the first test, the subject's class can be determined but that might change after one or more additional tests. Once enough tests have been performed with different parameter values, the results clearly reveal the strengths and weaknesses of the subject.
  • FIG. 8 is a flowchart of exemplary steps in a method 600 of assessing hearing characteristics of a subject.
  • the method can include, after the start at block 602, determining a hearing capability of the subject based on his responses to a series of sounds presented to him at block 604. Each sound corresponds to the presence, absence or irrelevance of a predetermined plurality of features.
  • the method 600 continues at block 606 by assigning the subject to one of a predetermined plurality of classes based upon the responses of the subject. Each of the plurality of classes is derived from hearing tests performed on a plurality of other subjects.
  • the method 600 can include, at block 608, setting one or more; parameters of a hearing-enhancement device based on the class to which the subject is assigned.
  • the method 600 illustratively concludes at block 610.
  • the step of determining hearing capability of a subject at block 604 can include identifying one or more of the plurality of features as contributing more than other of the plurality of features to a failure of the subject to correctly respond to the presentment of one or more of the series of sounds.
  • a failure to correctly respond to a particular one of the series of sounds more particularly, can define a feature error with respect to the one or more features corresponding to that particular one of the series of sounds.
  • the method 600 can further include measuring the performance of the subject based on a computed mean of feature errors.
  • the computed mean of feature errors can, moreover, equal a weighted mean, and the method can further comprise computing the weighted mean, ⁇ , to be equal to
  • W 1 is a weight assigned to the i th feature of the plurality of features and n t is the number of feature errors with respect to the i th feature.
  • the step of assigning the subject to one of the predetermined plurality of classes at block 606 can include computing a weighted contribution of each feature.
  • the weighted contribution of a feature can quantitatively measure the contribution that the feature makes to the computed mean of feature errors.
  • computing the weighted contribution of a feature can comprise computing a value equal to
  • the software may be configured to run on any computer or workstation such as a PC or PC-compatible machine, an Apple Macintosh, a Sun workstation, etc.
  • any device can be used as long as it is able to perform all of the functions and capabilities described herein.
  • the particular type of computer or workstation is not central to the invention, nor is the configuration, location, or design of the database, which may be flat-file, relational, or object-oriented, and may include one or more physical and/or logical components.
  • the servers may include a network interface continuously connected to the network, and thus support numerous geographically dispersed users and applications.
  • the network interface and the other internal components of the servers intercommunicate over a main bi-directional bus.
  • the main sequence of instructions effectuating the functions of the invention and facilitating interaction among clients, servers and a network can reside on a mass-storage device (such as a hard disk or optical storage unit) as well as in a main system memory during operation. Execution of these instructions and effectuation of the functions of the invention is accomplished by a central-processing unit ("CPU").
  • CPU central-processing unit
  • a group of functional modules that control the operation of the CPU and effectuate the operations of the invention as described above can be located in system memory (on the server or on a separate machine, as desired).
  • An operating system directs the execution of low- level, basic system functions such as memory allocation, file management, and operation of mass storage devices.
  • a control block implemented as a series of stored instructions, responds to client-originated access requests by retrieving the user-specific profile and applying the one or more rules as described above.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Public Health (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Veterinary Medicine (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • Surgery (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Evolutionary Computation (AREA)
  • Signal Processing (AREA)
  • Psychiatry (AREA)
  • Physiology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Otolaryngology (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

A computer-implemented method for generating a test set of test stimuli for testing a subject using a computer system may be used to reduce test time. The method includes selecting one or more features from among a plurality of features, wherein a measurable effect on the subject for each feature selected exceeds a threshold. Based on the selected features, one or more sample classes are generated. Test stimuli from each sample class are selected for presentation to the subject. Selection of the test stimuli includes choosing each stimulus from each sample class to present to the subject. These chosen stimuli make up the test stimuli. For each sample class that includes more than one stimulus, the method includes deselecting at least one stimulus from the test stimuli to reduce the test set size. Patient classification and the use of noise stimuli during testing may further reduce testing time.

Description

SYSTEM AND METHODS FOR REDUCING PERCEPTUAL DEVICE OPTIMIZATION TIME
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application claims priority to U.S. Patent Application Serial No. 12/201,492, filed August 29, 2008, entitled "System and Methods for Creating Reduced Test Sets Used in Assessing Subject Response to Stimuli"; U.S. Patent Application Serial No. 12/201,598, filed August 29, 2008, entitled "System and Methods of Subject Classification Based on Assessed Hearing Capabilities"; and to U.S. Provisional Patent Application No. 61/164,452, filed March 29, 2009, entitled "Systems and Methods for Determining the Nature of Hearing Weakness"; the disclosures of which are hereby incorporated by reference herein in their entireties.
FIELD OF THE INVENTION
[0002] The present invention is related to the field of hearing testing, and more particularly, to techniques for more efficiently assessing hearing capabilities and tuning hearing-enhancement and audio devices.
BACKGROUND OF THE INVENTION
[0003] Subject testing arises in a wide variety of contexts and ranges from posing a battery of test questions to eliciting a subject's response to different types of stimuli. A prime example of the latter type of testing is the technique used to fit a hearing-impaired subject, or patient, with a cochlear implant system. Once such a system is implanted, as with many other types of digital hearing-enhancement systems, a suitable speech coding and mapping strategy must be selected to enhance the performance of the system for day-to-day operation. The mapping strategy pertains to an adjustment of parameters corresponding to one or more independent channels of a multi-channel cochlear implant or other hearing-enhancement system. Selection of each strategy typically occurs over an introductory period of approximately six or seven weeks, during which the hearing-enhancement system is tuned for the particular patient. During this tuning period, users of such systems are asked to provide feedback on how they feel the device is performing. [0004] More particularly, to create a mapping for a speech processor, an audiologist first determines the electrical dynamic range for each electrode or sensor used. The programming system delivers an electrical current through the hearing-enhancement system to each electrode in order to obtain the electrical threshold (T-level) and comfort or "max" level (C-level) measures defined by a system's manufacturer. T-level, or minimum stimulation level, is the minimum electrical current capable of producing an auditory sensation in the user 100 percent of the time. The C-level is the loudest level of signal to which a user can listen comfortably for a long period of time.
[0005] A speech processor then is programmed or "mapped" using one of several encoding strategies so that the electrical current delivered to the implant will be within this measured dynamic range; that is between the T- and C-levels. After T- and C-levels are established and the mapping is created, the microphone is activated so that the patient is able to hear speech and other sounds. From that point onwards, the tuning process continues as a traditional hearing test. Hearing-enhancement device users are asked to listen to tones of varying frequencies and amplitudes. The gain of each channel can be further altered within the established threshold ranges such that the patient is able to hear various tones of varying amplitudes and frequencies reasonably well.
[0006] Not surprisingly, fitting and tuning a hearing-enhancement system of any type so as to meet the needs of a particular patient is typically quite costly and very time consuming, both from the perspective of the hearing-impaired patient and the audiologist. The functions of such a system are regulated by a large number of parameters, values for each of which typically must be determined so as to tune the system to provide optimal performance for the particular patient. In order to do so, the patient typically must be thoroughly tested with respect to each of set of parameter values. The number of tests generally increases exponentially as the number of system parameters increases. Additionally, the testing environment itself can be a factor when testing a user and setting system parameters. Often, in an "ideal" testing environment (i.e., one having little background noise), persons with mild to moderate weakness in perceptual ability may be able to overcome their weakness by increased concentration. However, any listener will fail to distinguish between a pair of phonemes in sufficient noise. There is a noise threshold for each pair of phonemes above which a listener fails to distinguish a particular sound or a particular class of sounds (such as classes based on the "distinctive feature theory" or "landmark theory"). If a person's noise level for failure is less than a particular threshold, that person may be deemed to have a hearing problem. Otherwise, they may be assumed to have acceptable perceptive ability for that pair of phonemes/class of sounds. By determining a user's noise level for failure, perceptual devices, such as cochlear implants or hearing aids, may be tuned to improve a user's hearing across a broad range of frequencies and/or noise volume levels. Testing in noise effectively and quickly exposes a person's hearing weakness, thus reducing overall test time and providing a more accurate model of their hearing weakness.
[0007] More generally, the problems inherent in testing a hearing-impaired patient so as to optimally set the system parameters values for a hearing-enhancement system arise with various other types of qualitative or quantitative tests in which a subject has to respond to posed test questions or physical stimuli. Many testing techniques, such as those utilized for tuning a hearing-enhancement system as described above, require a substantial investment of time and effort. Accordingly, there is a need for a technique that in such contexts is able to reduce testing time without compromising the quality of the testing performed.
SUMMARY OF THE INVENTION
[0008] With the foregoing background in mind, it is therefore a feature of the invention to provide systems and method for reducing time and effort expended in testing a subject without compromising the quality of the results yielded from the testing. One aspect of the invention is to provide system and methods that utilize techniques for identifying the raw perceptual weaknesses of a subject in a considerably shortened time frame as compared to conventional techniques.
[0009] In one aspect, the invention relates to a computer-implemented method for generating a test set having test stimuli for testing a subject using a computer system having logic-based processing circuitry, the method including: selecting one or more features from among a plurality of features, wherein a measurable effect on the subject for each feature selected exceeds a predetermined threshold; based on the selected features, generating one or more sample classes; selecting test stimuli from each sample class for presenting to the subject, wherein the selecting step includes: initially choosing each stimulus from each sample class to present to the subject, wherein the chosen stimuli include the test stimuli; and for each sample class that has more than one stimulus, subsequently deselecting at least one stimulus from the - A -
test stimuli. In an embodiment of the above aspect, at least one of the deselected test stimuli is deselected randomly. In another embodiment, the method further includes deselecting at least one stimulus associated with a non-selected feature to present to the subject. In still another embodiment, deselecting at least one stimulus associated with a non-selected feature is repeated iteratively. In yet another embodiment, iteratively deselecting at least one stimulus associated with a non-selected feature includes deselecting one less stimulus from a class including more than one stimulus.
[0010] In another embodiment of the above aspect, iteratively deselecting at least one stimulus associated with a non-selected feature includes randomly deselecting a stimulus. In another embodiment, the method further includes initially presenting each of the test stimuli prior to selecting from among the plurality of features associated with each of the test stimuli. In still another embodiment, selecting from among the plurality of features includes computing for a feature a product equal to a value assigned to the feature times a number of stimuli influenced by the feature. In yet another embodiment, the method further includes determining a measure of significance of each sample class on assessing the hearing capability of the patient. In another embodiment, the measure of significance of the ith class is equal to
Figure imgf000005_0001
e where nt is the number of stimuli contained in the i th class, andpy is an empirically determined value based upon an assessment of patient responses to administered test stimuli. In still another embodiment, the method further includes determining the number of times to present different test stimuli to the patient based upon a computed measure of patient hearing weakness, the measure, su being defined as
s, = k x-
wherein ei is a number of test stimuli selected from the i ■th class and erroneously recognized by the patient, wherein c is a number of test stimuli contained in the ith class, and wherein k is a predetermined constant. [0011] In another embodiment of the above aspect, the method further includes presenting the test stimuli to the subject. In another embodiment, the presenting step further includes: selecting at least two test stimuli; transmitting a noise stimulus to the subject; transmitting the two test stimuli to the subject; and receiving at least one associated user response to the two test stimuli, wherein the associated user response is based at least in part on the test stimuli. In still another embodiment, the method includes determining a relative weakness of a feature set for the subject, wherein the relative weakness is based at least in part on the noise stimulus and the two test stimuli. In yet another embodiment, the method includes adjusting a volume of the noise stimuli from the first volume to a second volume; and repeating the noise transmitting step, the test stimuli transmitting step, and the receiving step. In another embodiment, the adjusting step is based at least in part on the associated user response. In another embodiment, the first volume is greater than the second volume. In another embodiment, the associated user response is a non-response.
[0012] In another embodiment of the above aspect, the at least two test stimuli are a phoneme pair. In another embodiment, the noise stimulus has a temporal structure and the test stimuli have a temporal structure, and the noise temporal structure substantially matches the test temporal structure. In yet another embodiment, the transmitting step and the receiving step together are a sequential testing step, and the sequential testing step includes transmitting a first test stimulus, then receiving a first associated user response, then transmitting a second test stimulus, then receiving a second associated user response. In still another embodiment, the method includes determining a threshold noise volume. In another embodiment, the threshold noise volume is the first volume.
[0013] In another aspect, the invention relates to a computer-implemented method of testing a hearing-impaired patient using a computer system including logic-based processing circuitry, the method including the steps of: audibly presenting a plurality of phonemes to the patient, wherein each phoneme is selected from one of a plurality of phoneme sets corresponding to a predetermined feature selected for testing a hearing capability of the patient, wherein the plurality of phonemes are test stimuli; based on a response of the patient to each audibly presented phoneme, generating a first assessment of the hearing capability of the patient; and audibly presenting a second plurality of phonemes to the patient and generating a second assessment of the hearing capability of the patient based on patient response, wherein the second plurality of phonemes is selected by deselecting from the test stimuli at least one phoneme from each phoneme set that contains more than one phoneme, wherein at least one of the deselected phonemes is deselected randomly. In an embodiment of the above aspect, the method further includes presenting a third plurality of phonemes, wherein the third plurality of phonemes is selected by deselecting from the test stimuli at least two phonemes from each phoneme set that contains more than one phoneme, wherein at least one of the deselected phonemes is deselected randomly. In another embodiment the method includes adjusting at least one operational parameter of a hearing-enhancement device based at least in part on a hearing performance of the subject with respect to different parameter values. [0014] In another aspect, the invention relates to a computer-based system for generating a test set of test stimuli for testing a subject, the system including: at least one processor; an electronic memory having stored therein electronic data representing a plurality of features for testing a particular capability of the subject; a feature-selecting module configured to execute on the at least one processor for selecting one or more features from among the plurality of features, wherein a measurable effect on the subject of each feature selected exceeds a predetermined threshold; and a stimulus-selecting module configured to execute on the at least one processor for generating one or more classes of stimuli based upon the selected feature, and selecting at least one stimulus from each class for presenting to the subject; wherein the stimulus-selecting module is further configured to initially choose each stimulus from each class to present to the subject wherein the chosen stimuli include the test stimuli, and subsequently deselecting from the test set at least one stimulus from each class that includes more than one stimulus, wherein at least one of the deselected test stimuli is deselected randomly by the stimulus- selecting module. In an embodiment, the stimulus- selecting module is further configured to deselect at least one stimulus associated with a non-selected feature to present to the subject. In another embodiment, stimulus-selecting module iteratively deselects at least one stimulus associated with a non-selected feature. In yet another embodiment, iteratively deselecting at least one stimulus associated with a non-selected feature includes deselecting one less stimulus from a class having more than one stimulus. In still another embodiment, iteratively deselecting at least one stimulus associated with a non-selected feature includes randomly deselecting a stimulus. [0015] In another embodiment of the above aspect, the system is configured to initially present each of the predetermined number of stimuli prior to selecting from among the plurality of features associated with each of the predetermined number of stimuli. In another embodiment, the feature- selecting module is configured to select from among the plurality of features by computing for a feature a product equal to a value assigned to the feature times a number of stimuli influenced by the feature. In yet another embodiment, the system includes a significance-determining module for determining a measure of significance of each class of stimuli on assessing the hearing capability of the patient. In still another embodiment the measure of significance is equal to
J. [θ "' o lfth"e'r>w0ise where n, is the number of stimuli contained in the ith class, andpy is an empirically determined value based upon an assessment of patient responses to administered stimuli. In another embodiment, the system includes a stimulus-presentment module for determining a number of times to present different stimuli to the patient based upon a computed measure of patient hearing weakness, the measure, Si, defined as
Figure imgf000008_0001
wherein et is a number of stimuli selected from the ith class and erroneously recognized by the patient, wherein c is a number of stimuli contained in the ith class, and where k is a predetermined constant. [0016] In another aspect, the invention relates to a computer-based system of testing a hearing- impaired patient with a test set, the system including: an audio unit for audibly presenting a plurality of phonemes to the patient, wherein each phoneme is selected from one of a plurality of phoneme sets corresponding to a predetermined feature selected for testing a hearing capability of the patient, wherein the plurality of phonemes include the test set; and a testing unit including at least one processor for generating a first assessment of the hearing capability of the patient based upon a response of the patient to each audibly presented phoneme; wherein the system audibly presents a second plurality of phonemes to the patient and generates a second assessment of the hearing capability of the patient based upon patient response, wherein the second plurality of phonemes is selected by the testing unit deselecting from the test set at least one phoneme from each phoneme set that contains more than one phoneme, wherein at least one of the deselected phonemes is deselected randomly. In another aspect, the invention relates to a computer-readable medium having embedded therein computer-executable code that, when downloaded and executed by a computer system, causes the computer system to: select one or more features from among a plurality of features, wherein a measurable effect on the subject of each feature selected exceeds a predetermined threshold; based on the selected features, generate one or more classes of stimuli; and select test stimuli from each class for presenting to the subject, wherein the selection of test stimuli includes initially choosing each stimulus from each class to present to the subject, and subsequently deselecting from the test stimuli at least one stimulus from each class that includes more than one stimulus, wherein at least one of the deselected phonemes is deselected randomly. [0017] In another aspect, the invention relates to a computer-implemented method for generating a test set for testing a subject using a computer system including logic-based processing circuitry, the method including: selecting one or more features from among a plurality of features, the selecting including selecting one or more feature having a measurable effect on the subject, as determined based on a predetermined threshold; based on the selected features, generating one or more classes of stimuli; and selecting stimuli from one or more of said classes for presenting to the subject. In another aspect, the invention relates to a computer-implemented method for generating a test set for testing a subject using a computer system having logic-based processing circuitry, the method including: selecting one or more features from among a plurality of features, wherein a measurable effect on the subject for each feature selected exceeds a predetermined threshold; based on the selected features, generating one or more classes of stimuli; and selecting stimuli from one or more of said classes for presenting to the subject, wherein selecting includes choosing a stimulus from a class to present to the subject wherein the chosen stimuli includes the test set, and subsequently deselecting at least one stimulus from the test set from each class that includes more than one stimulus. [0018] In another aspect the invention relates to a method of assessing hearing characteristics of a subject, the method including: determining a hearing capability of the subject based on responses of the subject to a series of sounds presented to the subject, each sound corresponding to a presence, absence or irrelevance of a predetermined plurality of features; and assigning the subject to one of a predetermined plurality of classes based upon the responses of the subject, each of the plurality of classes being derived from hearing tests performed on a plurality of other subjects. In an embodiment of the above aspect, the method includes setting one or more parameter values of a hearing-enhancement device based on the class to which the subject is assigned. In another embodiment, the step of determining hearing capability includes identifying one or more of the plurality of features as contributing more than other of the plurality of features to a failure of the subject to correctly respond to the presentment of one or more, of the series of sounds. In yet another embodiment, a failure to correctly respond to a particular one of the series of sounds defines a feature error with respect to the one or more features corresponding to that particular one of the series of sounds, and further including generating a performance measure for the subject based upon a computed mean of feature errors. In still another embodiment, the computed mean of feature errors equals a weighted mean, and further including computing the weighted mean, ξ, to be equal to
N
WΛ ξ = ^ N
∑vn ι=l where W1, is a weight assigned to the ith feature of the plurality of features and nτ is the number of feature errors with respect to the ith feature. In another embodiment, the step of assigning the subject to one of the predetermined plurality of classes includes computing a weighted contribution of each feature, the weighted contribution of a feature quantitatively measuring the contribution that the feature makes to the computed mean of feature errors. In another embodiment of the above aspect, computing the weighted contribution of a feature includes computing a value equal to
Contribution{fi ) = -^ ξ wherein Contributionifi) is the weighted contribution of the ith feature. [0019] In another aspect, the invention relates to a system for tuning a hearing- enhancement device, the system including: a subject interface for rendering a series of sounds to a subject and for receiving from the subject a response to each of the sounds rendered, each sound corresponding to one or more features belonging to a predetermined plurality of features; and a processing unit communicatively linked, to said subject interface, the processing unit having: a hearing-capability module for determining a hearing capability of the subject based on the received responses of the subject to the series of sounds rendered; a class- assigning module for assigning the subject to one of a predetermined plurality of classes based upon the received responses, each of the plurality of classes being derived from hearing tests performed on a plurality of other subjects; and a tuning module for setting one or more parameters of a hearing-enhancement device based on the class to which the subject is assigned. In another embodiment of the above aspect, the hearing-capability module is configured to determine hearing capability of the subject by identifying one or more of the plurality of features as contributing more than other of the plurality of features to a failure of the subject to correctly respond to the presentment of one or more of the series of sounds. In another embodiment, a failure to correctly respond to a particular one of the series of sounds defines a feature error with respect to the one or more features corresponding to that particular one: of the series of sounds, arid wherein the hearing-capability module is further configured to measure a hearing performance of the subject, based on a computed mean of feature errors. In yet another embodiment, the computed mean of feature errors equals a weighted mean, and wherein the hearing-capability module is configured to compute the weighted mean, ξ, to be equal to
Figure imgf000011_0001
where W1 is a weight assigned to the ith feature of the plurality of features and nt is the number of feature errors with respect to the ith feature. In still another embodiment, the class-assigning module is configured to assign the subject to one of the predetermined plurality of classes by computing a weighted contribution of each feature, the weighted contribution of a feature quantitatively measuring the contribution that the feature makes to the computed mean of feature errors. In another embodiment, the class-assigning module computes the weighted contribution of a feature by determining a value equal to
Contribution(fι ) = ^y-
wherein Contributionifi) is the weighted contribution of the ith feature. [0020] In another aspect, the invention relates to a computer-readable storage medium in which computer-readable code is embedded, the computer-readable code configured to cause a computing system to perform the following steps when loaded on and executed by the computing system: determine a hearing capability of a subject based on responses; of the. subject to a series of sounds presented to the subject, each sound corresponding to a presence, absence or irrelevance of a predetermined plurality of features; and assign the subject to one of a predetermined plurality of classes based on his responses, each of the plurality of classes being derived from hearing tests performed on a plurality of other subjects. In an embodiment of the above aspect, the storage medium includes: computer- readable code for causing the computing system to set one; or more parameter values of a hearing-enhancement device based upon the class to which the subject is assigned. In another embodiment, the step of determining hearing capability includes identifying one or more of the plurality of features as contributing more than other of the plurality of features to a failure of the subject to correctly respond to the presentment of one or more of the series of sounds. In yet another embodiment, a failure to correctly respond to a particular one of the series of sounds identifies a feature error with respect to the one or more features corresponding to that particular one of the series of sounds, and further includes measuring a hearing performance of the subject based on a computed mean of feature errors. In still another embodiment, the computed mean of feature errors equals a weighted mean, and further including computing the weighted mean, ξ, to be equal to
Figure imgf000012_0001
where W1 is a weight assigned to the ith feature of the plurality of features and nt is the number of feature errors with respect to the ith feature. In another embodiment, the step of assigning the subject to one of the predetermined plurality of classes includes computing a weighted contribution of each feature, the weighted contribution of a feature quantitatively measuring the contribution that the feature makes to the computed mean of feature errors. In another embodiment, the computer-readable storage medium of claim 59, wherein computing the weighted contribution of a feature includes computing a value equal to
Contήbution{fι ) = -^-
wherein Contribution(fi) is the weighted contribution of the ith feature.
[0021] In another aspect, the invention includes an article of manufacture having a computer-readable medium with computer-readable instructions embodied thereon for performing the methods described in the preceding paragraphs. In particular, the functionality of a method of the present invention may be embedded on a computer- readable medium, such as, but not limited to, a floppy disk, a hard disk, an optical disk, a magnetic tape, a PROM, an EPROM, CD-ROM, DVD-ROM or downloaded from a server. The functionality of the techniques may be embedded on the computer-readable medium in any number of computer- readable instructions, or languages such as, for example, FORTRAN, PASCAL, C, C++, Java, PERL, LISP, JavaScript, C#, TcI, BASIC and assembly language. Further, the computer- readable instructions may, for example, be written in a script, macro, or functionally embedded in commercially available software (such as EXCEL or VISUAL BASIC).
BRIEF DESCRIPTION OF THE DRAWINGS
[0022] There are shown in the drawings, embodiments which are presently preferred. It is expressly noted, however, that the invention is not limited to the precise arrangements and instrumentalities shown in the drawings.
• FIG. 1 is a schematic diagram of a system for generating a test set for testing a subject, according to one embodiment of the invention.
• FIG. 2 is a schematic diagram of a system of testing a hearing-impaired patient, according to another embodiment of the invention.
• FIG. 3 is a flowchart of exemplary steps in a method for generating a test set for testing a subject, according to yet another embodiment of the invention.
• FIG. 4 is a flowchart of exemplary steps in a method for testing a subject utilizing noise stimuli, according to another embodiment of the invention. • FIG. 5 is a schematic diagram of an environment in which a system for tuning a hearing-enhancement device, according to one embodiment of the invention, can be utilized.
• FIG. 6 is a schematic diagram of another environment in which a system for tuning a hearing-enhancement device, according to a different embodiment of the invention, can be utilized.
• FIG. 7 is a more detailed schematic view of a system for tuning a hearing-enhancement device, according to one embodiment of the invention.
• FIG. 8 is a flowchart of exemplary steps in a method of testing a hearing-impaired subject, according to still another embodiment of the invention.
DETAILED DESCRIPTION
[0023] Referring initially to FIG. 1, a computer-based system 100 for generating a test set for testing a subject, according to one embodiment, is schematically illustrated. The system 100 illustratively includes one or more processors 102 comprising registers, logic gates, and other data processing circuitry (not explicitly shown) for executing processor-executable instructions. The system 100 also illustratively includes an electronic memory 104 for electronically storing processor-executable instructions and data.
[0024] Additionally, the system 100 illustratively includes a feature-selecting module 106, the operative features of which are described more particularly below and which is communicatively linked to the one or more processors 102. The system 100 further illustratively includes a stimulus-selecting module 108, the operative features of which are also described more particularly below. Like the feature- selecting module 106, the stimulus- selecting module 108 is communicatively linked to the one or more processors 102.
[0025] Accordingly, each of the feature- selecting module 106 and the stimulus- selecting module 108 can be implemented in a combination of the processing circuitry of the one or more processors and processor-executable instructions that, when loaded to and executed by, one or more processors performs the operations, procedures, and functions described herein. In an alternate embodiment, however, one or both the feature-selecting module 106 and the stimulus-selecting module 108 can be implemented in dedicated hardwired circuitry configured to perform the same operations, procedures and functions. Ones possessing the ordinary skills in the art will also recognize that a combination of processing circuitry, hardware and/or firmware may be utilized to implement the feature-selecting module 106 and the stimulus- selecting module 108.
[0026] Operatively, the system 100 stores in the memory 104 electronic data representing a plurality of features for testing a particular capability of the subject. In a preferred embodiment, the feature- selecting module 106 executes on the one or more processors and selects one or more of the features from among the plurality of features. As described more particularly below, the feature-selecting module 106 is configured to select those features that have a significant effect on the subject; that is, each feature selected exceeds a predetermined threshold. The stimulus- selecting module 108 executing on the one or more processors generates one or more classes of stimuli based on the selected feature. As used herein, a stimulus is any signal, question, or other other-response eliciting element that is conveyed to a subject to elicit one or more responses that can be used to determine a capability, condition, or attribute of the subject. Stimuli can be grouped into distinct classes corresponding to one or more features, wherein the features are those attributes associated with certain stimuli. As described herein, the subject's ability to discern certain features when presented with a corresponding stimulus indicates a particular capability or characteristic of the subject.
[0027] Subsequently, or simultaneously, in a preferred embodiment, as each class is generated, the stimulus- selecting module 108 selects at least one stimulus from each class for presenting to the subject. The stimulus-selecting module 108 is further configured to initially choose each stimulus from each class to present to the subject, and to subsequently select at least one less from each class that comprises more than one stimulus. According to this embodiment, at least one of the stimuli subsequently selected from each class that comprises more than one stimulus is selected randomly by the stimulus-selecting module 108. [0028] As is described below, a stimulus-selecting module may be presented with fewer than all stimuli if the tester determines a priori that certain stimuli are of lesser significance. Furthermore, the tester may select a stimulus from less than all classes for presenting to a subject if time saving is desired. However, either of these approaches to limiting stimuli or class presentation may compromise the quality of testing. [0029] Stimuli can be used to test particular capabilities of a subject, such as a patient. Specific stimuli can be associated with particular features that characterize the particular capabilities. Thus, generally, the purpose of eliciting a response of the subject to the stimuli is to identify features with respect to which the subject is weak. For example, in the context of testing hearing, two distinctive feature sets have been proposed. The first is based on the articulatory positions underlying the production of speech sounds. The other is based on the acoustic properties of various speech sounds. These properties describe a small set of contrastive acoustic properties that are perceptually relevant for the discrimination of pairs of speech sounds. More particularly, as will be readily understood by one of ordinary skill, the different distinctive features and their potential acoustic correlates can be broadly grouped into three categories: fundamental source features, secondary consonantal source features, and resonance features.
[0030] It should be noted that these are not the only ways in which various features can be categorized. There can also be additional features, or groups of features that more optimally describe different tasks. For example, with respect to a different language such as Mandarin using additional features, albeit in accordance with the process described herein, the results can be further enhanced.
[0031] The fundamental source features can be further characterized on the basis of whether the speech sounds are vocalic or non-vocalic. Vocalic speech corresponds to speech sounds associated with vowels. Accordingly, such speech sounds correspond to a single periodic source, the onset of the speech not being abrupt; otherwise the speech sound can be characterized as non-vocalic. The fundamental source features also can be characterized on the basis of whether the speech sounds are consonantal or non-consonantal. Consonantal speech sounds correspond to sounds associated with consonants. Such speech sounds are characterized by the presence of zeros in the associated spectrum of the sounds.
[0032] The secondary consonantal source features can be further characterized on the basis of whether the speech sounds are interrupted or continuant. Continuant speech sounds, are also characterized as semi- vowels, because of their similar sound quality. There is little or no friction with continuant speech sounds as the air passes out freely through the mouth of the speaker. A continuant speech sound is produced with an incomplete closure of the vocal tract. Interrupted speech sounds, by contrast, end abruptly. [0033] The secondary consonantal features can also be characterized on the basis of whether the speech sounds are checked or unchecked. Checked speech sounds, typified by some Far Eastern and African languages, are characterized by abrupt termination as opposed to gradual decay, whereas unchecked speech sounds are characterized by gradual decay. Additionally, secondary consonantal features can be characterized as strident or mellow. The former typically has an irregular waveform, whereas the latter typically has a smooth waveform. A secondary consonantal feature characterized as mellow also has a wider autocorrelation function relative to a corresponding normalized strident feature. Secondary consonantal features can also be classified according to whether the sound is voiced or voiceless.
[0034] The resonance features can be further characterized on the basis of whether the speech sound is compact or diffuse. A compact feature is associated with sound having a relative predominance of one centrally located format region, whereas a diffuse features implies sound having one or more non-central formats. The resonance features can also be characterized as grave or acute. Speech sounds that are characterized as grave are low- frequency dominant low frequency, whereas those characterized as acute are high-frequency dominant. Additionally, resonance features can be characterized as flat or plain, depending on whether the there is a downward shift of some or all formats, typically associated with vowels and a reduction in lip orifice of the speaker. [0035] The resonance features also can be further characterized as sharp or plain, the latter characterizing speech sounds whose second and/or higher formats rise. Moreover, resonance features can also be characterized as tense or lax, depending on the amount and duration of the energy of the sound. The resonance features also can be classified according to whether the speech sound is characterized as having a nasal format or a nasal murmur. The distinctive speech features and their potential acoustic correlates are further described in R. Jakobson, G. M. Fant, and M. Halle, PRELIMINARIES TO SPEECH ANALYSIS: THE DISTINCTIVE FEATURES AND THEIR CORRELATES (MIT Press, Cambridge; 1963), which is incorporated herein by reference in its entirety.
[0036] Whereas, in the context of testing hearing vocalic, consonantal, compact, and other of the described features are relevant, in other contexts different features will be found to be important. For example, in testing vision, features such as color intensity, object shape or size, and other visual features are important. Thus, although the system 100 is described primarily in the context of testing hearing, it will be readily apparent from the description herein, that the procedures, processes, and functions performed by the system pertain to other applications for testing a subject, such as a patient. [0037] In a preferred embodiment as described in details below, fourteen phonemes from the IOWA test [R. S. Tyler, J. P. Preece, and N. Tye-Murray, The Iowa Phoneme and Sentence Tests, Dept. of Otolaryngology-Head and Neck Surgery, The University of Iowa, Iowa City, IA, 1986] are used to represent a patient's hearing ability in terms of the features. These phonemes may be consonants or vowels. The testing might also be carried out using words, sentences, phrases, or other sounds as described in US Patent No. 7206463 to Krause, et al. which is incorporated herein by reference in its entirety.
[0038] In general, in any particular context, certain relevant features play a greater role than others in the process of the subject recognizing or otherwise responding to a certain set of stimuli. Moreover, in certain applications (e.g., hearing testing) the capability of the subject with respect to one feature is dependent on the subject's underlying capability with respect to another feature.
[0039] A subject weak with respect to a certain class of stimuli, characterized for example by an inability to recognize or respond to the stimuli, is typically weak with respect to one or more of the features associated with that class of stimuli. Generally, the size of test set of stimuli is combinatorial with respect to the number of features. Accordingly, it is advantageous to identify the features that play a significant role in determining the subject's hearing ability. The stimuli can be classified into a smaller number of classes based on the significant features. The number of classes of stimuli increases exponentially with the number of features; if the less significant features are not eliminated, many of the classes will be empty. [0040] Accordingly, the system 100 is configured to identify the vital features so as to reduce the number of stimuli classes and allow each class to contain a meaningful set of stimuli. Also, by emphasizing on the vital features, it is expected that the majority of a subject's weaknesses can be ascertained in a shorter period of time than is conventionally the case. Since the brain of a human being is generally adept at recognizing a stimulus from only partial information, it will be sufficient to rectify the weaknesses caused by the more influential features in, for example, the context of tuning a hearing-enhancement device. [0041] According to one embodiment, the feature-selecting module 106 is configured to determine whether the effect of a feature on a subject exceeds a predetermined threshold according to the following: fl if v Xm > θ [θ otherwise where θ is the predetermined threshold, V7- is the number of different values assumed by the ith feature for representing the effect of the test stimuli (e.g., +1 if the feature is present in the stimulus, -1 if it is absent, and 0 if the feature is irrelevant), m; is the number of different stimuli influenced by the Mh feature. Accordingly,/ is a quantitative measure of the role of the Mi feature. During testing in a preferred embodiment a threshold value of 7 produced excellent results. However, one of the ordinary skills in the art is to recognize that other threshold values may also be used to produce relevant phoneme discrimination. One of the ordinary skills in the art is to recognize that the feature selecting module can be configured differently such that the effect of the features can be less than a predetermined threshold.
[0042] Those features that, according to the determination made by the feature-selecting module 108, are assigned a value of 1, are considered to play a vital role. Thus, the stimuli can be classified based on the important features only.
[0043] The stimulus- selecting module 106, according to yet another embodiment, can be configured to determine the significance of the ith class of stimuli. In a test, the number of different stimuli chosen from each class is typically proportional to the significance of that class. More particularly, the stimulus- selecting module 106 can be configured to compute the following measure of significance of a class of stimuli:
Figure imgf000019_0001
where n, is the number of different stimuli in the ith class, andpy be the significance of the/'1 stimuli in the ith class on the stimuli-recognition ability of the subject. Therefore, significance of the ith class of stimuli on the recognition ability of a patient is given by the above calculation as performed by the stimulus selecting module 106. One of the ordinary skills in the art is to recognize that the stimuli selecting module can be configured differently to compute a measure of significance.
[0044] The generation of a test set (i.e., the selected stimuli), as determined by the system 100 according to one embodiment, then depends on three factors: first, the significance of a class of stimuli, as discussed above; second, the weakness of the subject with respect to particular features (in many settings, the weakness becomes apparent within a few tests, and thus the subsequent presentation of stimuli becomes dependent on the weaknesses); and third, the mitigation of testing error, which according to the present invention is achieved by randomized stimuli selection, as described more particularly below. [0045] With respect to the second factor, the stimulus- selecting module 108 can be configured to quantitatively assess the weakness with respect to one or more features. Specifically, the stimulus-selecting module 108 can be configured to compute the following:
e. s, = k x- (in)
Σ', where et is the number of erroneously recognized stimuli from the ith class and k is a constant. One of the ordinary skills in the art is to recognize that the stimuli selecting module can be configured differently to assess the weakness with respect to one or more features.
[0046] In the preferred embodiment, the stimulus selecting module is utilized to reduce the number of classes and the number of stimuli. One of the ordinary skills in the art is to recognize that the number of stimuli could be reduced a priori based on perceived lower stimuli significance. However, this may result in compromise of the testing quality.
[0047] The operative features of the system 100 are now described in the specific context of generating a test set or selected stimuli for testing the hearing capabilities of a patient. The test set, in turn, can be utilized in fitting or "tuning" hearing-enhancement devices, such as hearing aids and cochlear implants, mobile telephones or wireless/Bluetooth devices, etc. As already noted, the functions of such a device are regulated by a large number of parameters, values for each of which must be determined so as to tune the device to provide optimal performance for a particular patient. Each parameter is assigned one of many values. Conventional techniques for determining the optimal set of parameter values for each patient is difficult and time consuming. A patient typically must be thoroughly tested in order to ascertain the parameter values that yield optimum device performance for the patient.
[0048] The goal of testing the hearing-impaired patient is to ascertain his raw hearing ability independent of context and background knowledge. During testing, a series of consonant phonemes can be presented, as stimuli, and the patient's response assessed so as to identify any weakness in his hearing. Different parameters of the hearing-enhancement device can be adjusted accordingly.
[0049] The operative features of the system 100 permit the patient to be tested according to an adaptive method, according to a particular embodiment. During testing, audible renderings of phonemes are presented to the patient, rather than words or phrases. The phonemes are selected from a set of fourteen consonant phonemes. In a preferred embodiment, consonant phonemes are preferably used rather than vowel phonemes, because the latter are typically too easily perceived by a patient and do not reveal sufficient information pertaining to the patient's hearing capability. As now described, the system enables the number of phonemes utilized in testing to be reduced, affording a significant savings in time and resources without compromising the quality of the testing performed.
[0050] The patient's strengths and weaknesses are assessed based on the patient's response to phonemes corresponding to the different features represented by each. As already noted, a phoneme is characterized by the presence, absence or irrelevance of a set of nine features: Vocalic, Consonantal, Compact, Grave, Flat, Nasal, Tense, Continuant, and Strident. As described more particularly below, the feature- selecting module 106 can be configured to operate on features arranged hierarchically. Those features higher in the hierarchy are those that potentially have a greater effect because the failure to perceive these features affects the perception of a greater number of phonemes. Thus, those features ranked higher in this hierarchy provide a more comprehensive measure of hearing loss as compared to words or phrases. Each phoneme can be associated with a corresponding percentage of proportional occurrence of the phoneme in the English language. (See, e.g., L. Shriberg and R. Kent, Linical Phonetics, Boston: Allyn & Bacon (2003), incorporated here in its entirety.)
[0051] The presence, absence, and irrelevance of a feature with respect to each phoneme can be can be denoted, respectively, as 1, -1, and 0. The fourteen consonant phonemes used in testing and the corresponding constituent features of each are: neme s Vocalic Cons. Compact Grave Flat Nasal Tense Cont. Strident n -1 1 -1 -1 0 1 0 0 0 t -1 1 -1 -1 0 -1 1 -1 0
S -1 1 -1 -1 0 -1 1 1 1 d -1 1 -1 -1 0 -1 -1 -1 0 k -1 1 -1 0 0 -1 1 -1 -1 m -1 1 -1 1 0 1 0 0 0
Z -1 1 -1 -1 0 -1 -1 1 1 b -1 1 -1 1 0 -1 -1 -1 0
P -1 1 -1 1 0 -1 1 -1 0
V -1 1 -1 1 0 -1 -1 1 0 f -1 1 -1 1 0 -1 1 1 0 g -1 1 1 0 0 -1 -1 -1 -1 sh -1 1 1 0 0 -1 1 1 0 j -1 1 1 0 0 -1 -1 -1 1
[0052] The features Vocalic and Consonantal remain the same with respect to all fourteen phonemes. The features Tense, Continuant, and Strident do not make a substantial difference to hearing ability, as has been verified empirically. Moreover, the feature Flat does not influence any of the fourteen phonemes. The following exemplary pseudo-code illustrates a procedure that can be implemented by the feature-selecting module 108 for creating a hierarchy of the nine features: This hierarchy is derived in R. Jakobson, G. Fant, and M. Halle, Preliminaries to Speech Analysis, Cambridge, MA: The MIT Press, 1963, which is incorporated herein in its entirety. Note that in the following exemplary pseudo-code, the consonantal sound "δ", referred to as the "voiced dental non-sibilant fricative" in the International Phonetic Alphabet, is utilized in the line reading 'If Mellow -> "/ δ /" as in THat.'
If sound is Vocalic
Is it Consonantal or Non-consonantal If Consonantal -> /1/ If Non-consonantal: Is it Compact or Diffuse
If Compact
Is it Grave or Acute If Grave
Is it Flat or Plain If Flat -> /o/
If Plain -> /a/ If Acute -> Id If Diffuse
Is it Grave or Acute
If Grave Is it Flat or Plain
If Flat -> IwI
If Plain -> neutral vowel "schewa" If Acute -> IiI If sound is Non-vocalic
Is it Consonantal or Non-consonantal If Non consonantal
Is it Tense or Lax
If Tense -> IhI If Lax ->
If Consonantal
Is it Compact or Diffuse If Diffuse
Is it Grave or Acute If Grave
Is it Nasal or Oral
If Nasal -> /m/ If Oral
Is it Tense or Lax If Tense
Is it Cont./lntr.
If Con. -> /f/ If lntr. -> /p/ If Lax Is it Cont./lntr.
If Con. -> /y/ If lntr. -* /b/ If Acute
Is it Nasal or Oral If Nasal -> /n/
If Oral
Is it Tense or Lax If Tense
Is it Continuant or Interrupted. If Continuant
Is it Strident or Mellow
If Strident -> /s/ If Mellow -> /θ/ as in THanks
If Interrupted. -> /t/ If Lax
Is it Continuant or Interrupted. If Continuant
Is it Strident or Mellow
If Strident -> ITJ
If Mellow -> "/ 5 /" as in THat If Interrupted. -> /d/
If Compact
Is it Nasal or Oral If Nasal -> /ng/
If Oral
Is it Tense or Lax If Tense
Is it Continuant or Interrupted
If Continuant -> /sh/ as in "SHoe"
If Interrupted
Is it Strident or Mellow
If Strident -> /ch/ as in "CHurch"
If Mellow -> IkI
If Lax
Is it Continuant or Interrupted If Continuant -> /zh/ as in
"meaSure"
If Interrupted
Is it Strident or Mellow
If Strident -> /dz/ as in "Judge"
If Mellow -> /g/
[0053] With reference to the earlier-described calculation performed by the stimulus- selecting module 106 to determine the significance of the ith class of stimuli, v(/th feature) = the number of different values assumed by the ith feature for representing the test stimuli. For example, Vocalic assumes only one value, '-1' for all the 14 phonemes. Since it fails to change among different values, v(Vocalic) = 0. Tense, for example, assumes three different values, '+1\ '-1', '0' for the 14 phonemes, and therefore, v(Tense) = 3. Thus, v(ith feature)=n-l, if n<2; otherwise v(ith feature) = n, where n is the number of values assumed by the ith feature.
[0054] The m(ith feature) = the number of different stimuli influenced by the ith feature. For example, Nasal can affect at most 4 phonemes; that is, perceiving Nasal incorrectly can affect at most 4 phonemes.
[0055] The m(ith feature) is dependent on the feature hierarchy, not just on the number of +1, -1 or 0 values assumed for the different phonemes. The feature hierarchy is a tree-like structure with the features occurring at the non-leaf nodes and the phonemes occurring at the leaf nodes. To reach a phoneme at a leaf node, one has to traverse the path (or branches) from the tree top (or root) down to the phoneme. One phoneme can be reached via exactly one path, so if a mistake occurs somewhere in the middle of the path (e.g., +1 is chosen for Nasal instead of-1), then it is impossible to reach that phoneme. More significant features, such as Vocalic and Consonantal, are higher up in the tree-like hierarchy, while the less significant features, such as Continuant and Strident, are at the lower levels.
[0056] At lower levels, the tree branches out. As a result, there are many non-leaf nodes at the lower levels that contain Continuant or Strident while there is only one node at the highest level with Vocalic. Therefore, getting Vocalic wrong results in all phonemes on one side of the entire tree being wrong. However, getting Strident wrong results in only a very few phonemes on one side of the tree starting from that non-leaf node being wrong. Thus m(Strident) is much less than m(Vocalic). In other words, m(ith feature) calculates the size of the largest number of phonemes on one side of the tree starting from the non-leaf node corresponding to the ith feature.
[0057] Bearing this in mind, the described calculations, in this example, yield: f(vocalic), m=14, v=0 -> f=0 f(cons.), m=14, v=0 -> f=0 f(compact), m=10, v=2 -> f=20 f (grave), m=5, v=3 -> f=15 f(flat), m=14, v=0 -> f=0 f(nasal), m=4, v=2 -> f=8 f(tense), m=2, v=3 -> f=6 f (cont) , m= 1 , v=3 -> f =3 f (strident), m=l, v=3 -> f=3
[0058] If the threshold is chosen to be 7, then Compact, Grave, and Nasal are the features selected.
[0059] Once these three features have been selected, the phonemes are classified based on the values assumed by these three features. Each of these features can assume three values:
+1, -1, 0. Thus, there can be 3 = 27 different combinations of the values assumed by the three features. Each unique combination of values of the selected features forms a class. So, the combination -1, 1, 1 is a distinct class and so is the combination -1, 1,-1. Most of these classes do not contain any phoneme. Only five of them do. For example, the class 1, 1, 1 does not contain any phoneme, the class -1, 1, .1 contains only 'm', while the class -1, 1, -1 contains {b, P, v, f}. [0060] Alternatively, if there is no concern regarding the perception of irrelevance of a feature, but there is with the presence or absence of the feature, then with the same three features, the minimal testing set will contain at most 23 = 8 phonemes. In this example, the possible combinations may be, therefore.
Compact Grave Nasal Phonemes representing
Combinations
1 1 1 —
1 1 -1 —
1 -1 1 —
1 -1 1 —
-1 1 1 m
-1 1 -1 {b, p, v, f}
-1 -1 1 n
-1 -1 -1 {t, s, d, z}
[0061] The remaining four phonemes {k, g, sh, j} have three important features as 1, 0, -1. Though that does not belong to any of the eight combinations sought, one of them can be used to test for <1 1 -l> or <l -1 -1>, especially given that there is no phoneme that has any of these two combinations. Thus, a minimal set of consonant phonemes contains five phonemes - m, n, and one each from the subsets {b, p, v, f}, {t, s, d, z}, and {k, g, sh, j}. The set is devoid of redundancy with respect to the features of interest.
[0062] The resulting test set is suitable for testing under ideal conditions. In practice, however, testing errors do occur. Thus, using the system 100 in the context of testing hearing, the hearing-impaired patient is tested with a few redundant phonemes in order to compensate for testing errors. Moreover, a patient may have hearing loss only in particular frequency ranges, which may not be evident from feature analysis. Thus, the testing uses redundant phonemes to mitigate these problems.
[0063] Additionally, testing under less than ideal conditions may also reduce test time and more quickly identify a patient's weakness in ability to distinguish phonemes. This may include testing the resulting test set of phonemes in varying levels of different kinds of noise. One proposed testing method includes testing with a prescribed set of phonemes in varying levels of different kinds of noise. Types of noise that may be utilized include white noise, pink noise, other broadband noises generated by filtering white noise and/or shaping the temporal structure of the noise envelop, environmental noises (sirens, alarms, machine noise, etc.), unintelligible/indistinguishable speech ("speech babble," as commonly encountered in a room containing a large number of people), as well as distorted speech such as that obtained through frequency and/or temporal compression, translation or any other manipulation, or by altering the phase of different components. The threshold level is first determined based on a level of phoneme or feature errors for a particular user. In certain embodiments of the method, test stimuli, such as a series or pairs of phonemes (e.g., 'b' and V, 'p' and 'f, 'm' and 'n', etc.) may be transmitted to a user, while the user is also exposed to the noise stimulus. The series or pairs may include phonemes having similar features, as well as unrelated features. The user is then asked to distinguish between the series or pairs of phonemes by selecting a proposed response, typing a response, speaking a response, etc. The background noise level may be varied until the patient fails to distinguish between the test stimuli. The amount of noise required to make the patient fail to distinguish the test stimuli provides an indication of the nature of weakness in the hearing ability of the patient. Alternatively or additionally, the speech perception task may be made more difficult by other factors, such as by systematically reducing the gain (output level), by adding additional tasks that require the listener's attention, etc.
[0064] In certain testing procedures, testing may be performed with each phoneme. In alternative embodiments, only certain pairs of phonemes may be utilized, thereby reducing the testing time. During testing, the noise stimulus is gradually reduced from the threshold level, and the phonemes are not previously identified to the patient. This prevents the patient from knowing what the testing entails, which may provide a more accurate test result. By beginning testing with the noise stimulus at the threshold level, the patient is unable to know the phonemes until he can actually perceive them. While testing with the noise stimulus beginning at the threshold level may be desirable, testing with a lower noise stimulus level may also be utilized. Various strategies for varying the noise may be employed so that the actual time for the entire test is less than the traditional methods of testing with the prescribed set of phonemes. These strategies include, but are not limited to, isolation of a break point of a single feature.
[0065] Testing systems may employ one type or several types of noise, as described briefly above. The nature of the noise may have a significant impact on the end results - for example, a phoneme pair intended to test perception of signals at high frequencies may necessitate the use of a noise stimulus with sufficient energy in the same frequency region. In contrast, other phoneme pairs may be tested more accurately with a noise having greater temporal variation. In certain embodiments, the noise stimulus has a temporal structure that substantially matches the temporal structure of the test stimuli. Further, the type of responses expected may also impact the accuracy of the testing. While it may be less complex to utilize a closed set of responses from which a user may choose, users may be able to guess the correct response, providing less accurate results. The ability to guess the response may necessitate multiple presentations to "average out" the guesses. Such multiple presentations may increase the test time. [0066] Whether tested under ideal conditions or utilizing background noise as described above, the phonemes are selected based on the patient's weaknesses. In the first test, all fourteen phonemes are presented, since initially there is no assessment of the patient's weaknesses. In the next two tests, eleven phonemes (three from each of the three sets described above) are presented. In the fourth test onwards, eight phonemes (two from each of the three sets) are presented. In the second and third tests, two phonemes from each set are chosen based on the patient's performance model, which captures the feature errors in the previous test. The remaining one phoneme from each set is chosen randomly from the remaining two phonemes. From the fourth test onwards, one phoneme from each set is chosen based on the patient's performance model, while the other phoneme is chosen randomly from the remaining three phonemes. The strategy is summarized as follows:
Phoneme Sets Test #l Tests #2 and #3 Test #4 and subsequent
{m} m m m
{n} n
{b, p, v, f} b, p, v, f choose 3 of choose 2 of which one which one randomly randomly
{t, s, d, z} t, s, d, z choose 3 of choose 2 of which one which one randomly randomly
{k, g, sh, j} k, g, sh, j choose 3 of choose 2 of which one which one randomly randomly
[0067] FIG. 2 is a schematic diagram of a computer-based system 200 for testing a hearing-impaired patient, according to another embodiment of the invention. The system 200 illustratively includes an audio unit 202, for audibly presenting a plurality of phonemes to the patient, wherein each phoneme is selected from one of a plurality of phoneme sets corresponding to a predetermined feature selected for testing a hearing capability of the patient. Additionally, if testing in noise is desired, the audio unit may present the selected noise stimuli. The system 200 further includes a testing unit 206 comprising at least one processor 208 for executing the procedures described above in generating assessments of the hearing capability of the patient based upon patient responses audibly-presented phonemes. [0068] The audio unit 202 can comprise, for example, a speaker, headphones, or other electromechanical transducer (not explicitly shown) for generating sound signals that can be played or otherwise rendered to the patient. In different embodiments, the audio unit 202 can be a hearing-enhancement device, a telephone, wireless phone, cellular phone, or the like.
[0069] The audio unit 202 can optionally include a microphone or other acoustical transducer for converting audible responses of the patient into electrical signals that are conveyed to the testing unit 206. The system 200 can optionally include a separate patient- response device 204, such as a hand-held push-button device, a keypad or the like that can be used by the patient in response to audibly-presented phonemes. The purpose of these different arrangements is to permit the system 200 to present the plurality of phonemes to which the patient responds so as to assess hearing capabilities of the patient. Optionally, the system 200 can also include a recorder 210 for recording the audible responses or signals conveyed to the testing unit 206 by the patient using the optional patient-response device 204.
[0070] Operatively, the testing unit 206 generates a first assessment of the hearing capability of the patient based upon a response of the patient to each audibly presented phoneme, according to the procedures described above. Additionally, the testing unit 206 may generate a predetermined noise stimuli if testing with background noise is desired. The system 200, subsequently, audibly presents a second plurality of phonemes to the patient and generates a second assessment of the hearing capability of the patient based upon patient response. The second plurality of phonemes is selected by the testing unit 206 choosing at least one less phoneme from each phoneme set that contains more than one phoneme, one of the phonemes selected from each phoneme set containing more than one phoneme being selected randomly by the testing unit. Again, background noise may be utilized during the entire test, or any portion thereof.
[0071] The system 200 optionally can include one or more databases 212 for storing the plurality of phonemes and noise stimuli. Although, the audio unit 202 is shown as communicatively linked directly, wirelessly or through a wire-line connection, with the testing unit 206, it will be readily apparent to one skilled in the relevant art that the system can be communicatively linked to the audio unit through one or more intermediate communication nodes of voice-based network or data communications network, such as the Internet. [0072] FIG. 3 illustrates certain method aspects of the invention. FIG. 3 is a flowchart of exemplary steps in a method 300 of generating a test set for testing a subject using a computer system comprising logic-based processing circuitry. The method 300 illustratively includes, after start at block 302, selecting one or more features from among a plurality of features at block 304. The selecting, more particularly comprises selecting one or more features having a measurable effect on the subject, as determined based on a predetermined threshold. The method 300 further includes generating one or more classes of stimuli based on the selected features at block 306. Additionally, the method includes selecting stimuli from one or more of the classes at block 308a. In parallel to step 308a, if testing is to be performed with noise stimulus, the appropriate noise stimulus is selected 308b. A noise stimulus that generally matches or approximates the temporal structure of the test stimuli may be desirable to accurately test the perceptual weakness of a subject. In an alternative embodiment of the method, a universal noise stimulus (e.g., speech babble) may be utilized in all testing, as opposed to a noise stimulus similar to the test stimuli. In such a case, the noise stimulus need not be selected in parallel with the test stimuli, as depicted in FIG. 3, but instead may be a predefined noise stimulus, regardless of test stimuli. The selected test stimuli and noise stimulus (if utilized) are then presented to the subject at block 310. The method illustratively concludes at block 312.
[0073] Using the methods described above to select the appropriate phonemes for testing, FIG. 4 depicts a procedure for testing a subject utilizing background noise 350. The procedure begins by selecting at least two test stimuli 352, in accordance with the methods described above. Thereafter, the noise stimulus (selected in Step 304a of FIG. 3) is transmitted to the subject 354. This noise stimulus may be transmitted or presented at a threshold volume, that is, the maximum volume that a subject may comfortable listen to. This threshold volume may be predetermined at set at the time of testing, or may be determined prior to presentation of the noise stimulus. Thereafter, the test stimuli is transmitted to the subject 356 and the testing system waits for a response. The response from the subject is received 358, either via a spoken sound, keyboard signals, or other forms described above. In certain embodiments, the subject response may be a non-response or a delayed response, thus signaling the testing system that the subject may be fatigued and that testing should be terminated. At this point, the method may enter a sequential testing loop 360, where a first stimulus is used to elicit a first response from the user, followed by a second stimulus, a second response, and so on. The processor receiving the user responses stores the various test stimuli and associated responses for later determination of subject perceptual weakness.
[0074] After a single stimuli/response procedure, or after one or more sequential testing loops, the system processor determines the relative perceptual weakness of the subject 362, in accordance with the methods described herein. At this point, the testing may conclude and the information obtained therefrom may be used to optimize or otherwise tune a hearing or other perceptual device. Exemplary tuning procedures and systems include those described in U.S. Patent No. 7,206,416, the disclosure of which is hereby incorporated by reference herein in its entirety. Alternatively, the procedure may continue. In one embodiment, the volume of the noise stimuli 364 may be adjusted and the testing procedure repeated 366. In certain cases, the new, adjusted noise stimuli volume may be increased or decreased, relative to the previously- tested volume. Adjustment of the noise volume may aid in determining the strengths and weaknesses of a subject, thus reducing overall test time.
[0075] In another embodiment, a method of generating a test set for testing a subject using a computer system comprising logic-based processing circuitry includes selecting one or more features from among a plurality of features, wherein a measurable effect on the subject of each feature selected exceeds a predetermined threshold. Additionally, method includes generating one or more classes of stimuli based on the selected features. The method further includes selecting stimuli from one or more of the classes for presenting to the subject, wherein selecting comprises choosing a stimulus from a class to present to the subject, and subsequently selecting at least one less from each class that comprises more than one stimulus.
[0076] Still another embodiment of a method for generating a test set for testing a subject using a computer system comprising logic-based processing circuitry includes selecting one or more features from among a plurality of features, wherein a measurable effect on the subject of each feature selected exceeds a predetermined threshold, and based on the selected features, generating one or more classes of stimuli. The method further includes selecting stimuli from each class for presenting to the subject, wherein selecting comprises initially choosing each stimulus from each class to present to the subject, and subsequently selecting at least one less from each class that comprises more than one stimulus, at least one of the stimuli subsequently selected from each class that comprises more than one stimulus being selected randomly. [0077] Another aspect of the invention is depicted with regard to FIGS. 5 and 8. FIG. 5 is a schematic view of an exemplary environment 400 in which system 402 for tuning a hearing- enhancement device, according to one; embodiment, can be utilized. The exemplary environment 400 illustratively comprises, in addition to the system 402, an audio unit 404 that delivers sound signals to the system. The audio unit 404 can comprise, for example, a speaker, headphones, or other electromechanical transducer (not explicitly shown) for generating sound signals in response to electrical signals that can be conveyed from the system 402 to the audio unit, the sounds being rendered then to a user of the audio unit. Indeed, in a particular embodiment, the audio unit 404 can be, the hearing-enhancement device that is to be tuned. In an alternative embodiment, however, the hearing-enhancement device can be a separate device that also connects to the system 402. [0078] The audio unit 404 can optionally include a microphone or other acoustical transducer for converting acoustic signals generated by the subject of the audio unit 404 into electrical signals that are conveyed to the system. The, exemplary environment 400 might additionally or alternately include a separate subject-response unit 406, such as a computer terminal for. presenting a graphical of other subject interface with which a subject interacts using a keyboard and/or computer mouse (neither explicitly shown). The purpose of these different arrangements, as described more particularly below, is to permit the system 402 to present to the subject a series of sounds to which the subject responds so as to assess hearing capabilities of the subject. [0079] According to a particular embodiment, the system 402 includes a subject interface 108 configured to communicatively link the system to the audio unit 404. If the exemplary environment 400 additionally or alternatively includes separate subject-response unit 406, then the subject interface 408 (or a corresponding one) also can be configured to communicatively link the system 402 to the subject-response unit. Thus, the subject interface 408 can be used in playing the series of sounds that are presented to the subject to a subject and for receiving, from the subject a response to each of the: sounds played. Each such sound, as described more particularly below, corresponds to the presence, absence or irrelevance of a predetermined plurality of features. Illustratively, the system 402 further includes a processing unit 410 for effecting the operative processes, procedures, and functions that are described more particularly below.
[0080] Optionally, the system 402 can also, include one or more databases; 412 for storing the plurality of features and/or sound signals, such as phonemes, words, etc., that correspond to the presence, absence or irrelevance of the plurality of features. Although, the system 402 is shown as communicatively linked directly, wirelessly or through a wire-line connection, with the audio unit 404, it will be readily apparent to one skilled in the relevant art that the system can be communicatively linked to the audio unit through one or more intermediate nodes.
[0081] FIG. 6 illustrates one such embodiment in which a system 502 for tuning a hearing- enhancement device is linked to an audio unit 504 through a data communications network 506, such as a local-area network (LAN), a wide-area network (WAN), or a plurality of connected network such as the Internet. Alternatively or additionally, the system 502 and audio unit 504 can be communicatively linked through a public- switched telephony network. As already noted above, a hearing-enhancement device being tuned can also serve as the audio unit that connects to the system, or alternatively, the hearing-enhancement device and audio unit can be separate devices. Thus, as illustrated with this embodiment, the system 502 can optionally connect separately to the audio unit 504 and to a hearing-enhancement device that is tuned by the system through the same network 506 or even a separate one.
[0082] Referring now to FIG. 7, a more detailed schematic view is provided of the system 402 depicted in FIG. 5, according to a particular embodiment. According to this embodiment, the processing unit 410 of the system 402 includes a hearing-capability module 414. The system 402, according to this embodiment, further illustratively includes a class-assigning module 416 communicatively linked to the hearing-capability module 414. Additionally the system illustratively includes a tuning module 418. One or more of the hearing-capability module 414, class-assigning module 416, and a tuning module 418 can be implemented in a combination of logic-based circuitry and processor-executable code. Thus; the processing unit 410, can be a general-purpose computer or application- specific computer having one or more processors implemented with registers, arithmetic-logic units, control units, and/or other logic- based circuitry. Alternatively, one or more of the hearing-capability module 414, class- assigning module 416, and a tuning module 418 can be implemented in dedicated, hardwired circuitry configured to operate cooperatively with other elements of the system 402.
[0083] Operatively, the subject interface 408 plays a series of sounds to the subject and, in response thereto, receives from the subject a response to each of the sounds played. Each sound, as already noted, corresponds to the presence, absence or irrelevance of a predetermined plurality of features. The hearing-capability module 414 of the processing unit 410 determines a hearing capability of the subject based on the received responses to the series of sounds played. The class-assigning module 416 assigns the subject to one of a predetermined plurality of classes based upon the received responses. Each of the plurality of classes consists of none, one or more subjects based on their hearing characteristics as assessed from the hearing tests performed on them. The tuning module 418 operates by setting one or more parameters of the hearing-enhancement device based on the device parameter settings of a similar subject in the class to which the subject is assigned. [0084] In a preferred embodiment, the series of sounds presented by the system 402 to the subject are phonemes. Thus, during testing, a series of phonemes are presented and after each presentment the subject responds, either audibly or using a subject-response unit as described above. Optionally, the system 402 can include a recording device (not shown) to record not only the subject's response but the response time of the subject. Based on the responses, the system 402 can incrementally build a model from which the strengths and weaknesses of the subject can be determined. This knowledge, in sum, can be used to determine parameter values for tuning the particular hearing-enhancement device, so that it performs at or sufficiently near to an optimal level. [0085] It is known that in natural language, a phoneme is the smallest unit of speech that distinguishes meaning. Words and sentences are a combination of phonemes in a particular order. According to a preferred embodiment, the system 402 is configured to present to the subject phonemes selected from a set of fourteen consonant phonemes (as distinct from words or sentences), the phonemes being those identified in the Iowa Medial Consonant Recognition Test. Vowel phonemes need not be utilized by the system, primarily because such are considered too easy to perceive and thus much less likely to be useful indicators of the nature of hearing capabilities or hearing loss.
[0086] Similar to that described above with regard to FIG. 1, the system 402 provides an analysis of the subject's strengths and weaknesses in terms of a predetermined set of features associated with each phoneme. Phoneme characterization, percentage of proportional occurrence in the English language, and feature hierarchical arrangement are described above. As shown in U.S. Patent No. 7,206,416 to Krause, et al., these features provide a more comprehensive measure of the subject's nature of hearing loss as compared to words or sentences. In a preferred embodiment, the system 402 is configured to classify subjects based on their performance in the three features: Compact, Grave, and Nasal.
[0087] An objective of the system 402 is to significantly reduce resource expenditures and time in testing. One way to achieve this is by using knowledge obtained through testing of other subjects or subjects in the past. The knowledge can be used to suggest parameter values for tuning, a hearing-enhancement device for a particular subject. Two subjects often have similar hearing characteristics, in which case it is highly probable that the optimum parameters values for one of the subjects are the same or very nearly so for the other. Using knowledge obtained from previously-tested subjects can thus contribute significantly to the goal of improving the effectiveness and efficiency of testing a subject and setting parameter values by suggesting those values most likely to be the optimal settings. Utilizing the suggested parameter values can reduce testing and tuning times significantly. [0088] A subject's performance in a test can be measured by the number of phonemes that the subject fails to perceive correctly. This, however, fails to capture the subject's strengths and weaknesses because many phonemes share similar features. For example, the phonemes T and 'p' differ only in one out of the nine features called Continuant. A person who fails to perceive 'p' due to an error in any feature other than Continuant will likely fail to perceive 'f as well. The converse is also true. Thus counting the number of phoneme errors is likely to be meaningless because feature errors give rise to phoneme errors. For the same reason, in order to reduce the phoneme errors, the system 402 is configured to concentrate instead on feature errors.
[0089] According to one embodiment, the hearing-capability module 414 is configured to determine hearing capability of the subject by identifying one or more of the plurality of features as contributing more than the others to a failure of the subject to correctly respond to the presentment of one or more of the series of sounds. More particularly, the hearing- capability module 414 can be configured to treat a failure to correctly respond to a particular one of the series of sounds as a feature error. Each feature error corresponds to a particular one of the series of sounds, which preferably are each phonemes.
[0090] Accordingly; the hearing-capability module 414 can be further configured to measure the hearing performance of the subject based on a computed mean of feature errors. Specifically, the computed mean of, feature errors can be a weighted mean, and the hearing- capability module 414 can be configured to compute the weighted mean, ξ, to be equal to
Figure imgf000036_0001
where W1 is a weight assigned to the ith feature of the plurality of features and nt is the number of feature errors with respect to the ith feature. [0091] An experimentally-determined set of weights for the nine above-listed features - Vocalic, Consonantal, Compact, Grave, Flat, Nasal, Tense, Continuant, and Strident - is {0.151785714, 0.151785714, 0.098214286, 0.0, 0.142857143, 0.125, 0.125, 0.0625}.
[0092] The system 402 can classify a subject based on the subject's strengths and weaknesses in perceiving phonemes. The objective is to classify subjects with similar hearing characteristics in the same class and subjects with different hearing characteristics in different classes. Two subjects are considered to have similar hearing characteristics if the same feature contributes more errors than another feature. As already noted, only three features - Compact, Grave, and Nasal - contribute significantly in determining the nature of hearing loss for a majority of subjects. The weights assigned to these features from the experimentally determined set are 0.142857143,0.098214286, and 0.142857143, respectively.
[0093] For each subject, the class-assigning module 416 can be configured to rank the three features based on their weighted contribution to the total weighted error. Thus for the ith feature,/^, among the plurality of features, the class-assigning module 416 can be configured to compute the weighted contribution to be
Contήbution(fι ) = -^ ξ (v) wherein Contribution(fi) is the weighted contribution of the ith feature.
[0094] Based on the ranking of this particular three-clement set of features, six classes are possible: <Compact, Grave, Nasal>
<Compact, Nasal, Grave> <Nasal, Compact, Grave> <Grave, Compact, Nasal> <Grave, Nasal, Compact> <Nasal, Grave, Compact> where <X, Y, Z> denotes the weighted contribution of errors from "the feature X is more than that of the feature Y, which is, in turn, more than that of the feature Z. Thus, for example, <Compact, Grave, Nasal> denotes the weighted contribution of errors from the feature Compact is more than that of the feature Grave, which is, in turn, more than that of the feature Nasal. Each of the other combinations similarly identifies: which feature contributes more to error than do the others among the plurality of features.
[0095] A subject belongs to one of the six classes. It is noted that when testing begins initially, the class to which the subject belongs is unknown. After the first test, the subject's class can be determined but that might change after one or more additional tests. Once enough tests have been performed with different parameter values, the results clearly reveal the strengths and weaknesses of the subject.
[0096] Membership of two subjects in the same class suggests that the same features contribute the most and least to their hearing weaknesses. Since features constitute phonemes, it is expected that both subjects will have similar phoneme errors and hence close optimal parameter values for the same hearing-enhancement device. Knowing the successful parameter values of a subject from a class thus rapidly suggests a good set of parameter values for another subject belonging to the same class, which can be determined as soon as the other subject's class can be determined with some level of certainty, since both of their optimal performances occur at nearby locations in the respective models. Accordingly, subject classification can significantly improve the efficacy and efficiency of testing hearing and tuning a hearing- enhancement device.
[0097] Certain method aspects of the invention are illustrated in FIG. 8. FIG. 8 is a flowchart of exemplary steps in a method 600 of assessing hearing characteristics of a subject. The method can include, after the start at block 602, determining a hearing capability of the subject based on his responses to a series of sounds presented to him at block 604. Each sound corresponds to the presence, absence or irrelevance of a predetermined plurality of features. The method 600 continues at block 606 by assigning the subject to one of a predetermined plurality of classes based upon the responses of the subject. Each of the plurality of classes is derived from hearing tests performed on a plurality of other subjects. Optionally, the method 600 can include, at block 608, setting one or more; parameters of a hearing-enhancement device based on the class to which the subject is assigned. The method 600 illustratively concludes at block 610.
[0098] According to one embodiment of the method 600, the step of determining hearing capability of a subject at block 604 can include identifying one or more of the plurality of features as contributing more than other of the plurality of features to a failure of the subject to correctly respond to the presentment of one or more of the series of sounds. A failure to correctly respond to a particular one of the series of sounds, more particularly, can define a feature error with respect to the one or more features corresponding to that particular one of the series of sounds. Thus, the method 600 can further include measuring the performance of the subject based on a computed mean of feature errors.
[0099] The computed mean of feature errors can, moreover, equal a weighted mean, and the method can further comprise computing the weighted mean, ξ, to be equal to
Figure imgf000039_0001
where W1, is a weight assigned to the ith feature of the plurality of features and nt is the number of feature errors with respect to the ith feature.
[0100] According to still another embodiment, the step of assigning the subject to one of the predetermined plurality of classes at block 606 can include computing a weighted contribution of each feature. The weighted contribution of a feature can quantitatively measure the contribution that the feature makes to the computed mean of feature errors. Moreover, computing the weighted contribution of a feature can comprise computing a value equal to
Contήbution(fι ) = -^ ξ (v) wherein Contribution^) is the weighted contribution of the ith feature.
[0101] In the embodiments described above, the software may be configured to run on any computer or workstation such as a PC or PC-compatible machine, an Apple Macintosh, a Sun workstation, etc. In general, any device can be used as long as it is able to perform all of the functions and capabilities described herein. The particular type of computer or workstation is not central to the invention, nor is the configuration, location, or design of the database, which may be flat-file, relational, or object-oriented, and may include one or more physical and/or logical components. [0102] The servers may include a network interface continuously connected to the network, and thus support numerous geographically dispersed users and applications. In a typical implementation, the network interface and the other internal components of the servers intercommunicate over a main bi-directional bus. The main sequence of instructions effectuating the functions of the invention and facilitating interaction among clients, servers and a network, can reside on a mass-storage device (such as a hard disk or optical storage unit) as well as in a main system memory during operation. Execution of these instructions and effectuation of the functions of the invention is accomplished by a central-processing unit ("CPU"). [0103] A group of functional modules that control the operation of the CPU and effectuate the operations of the invention as described above can be located in system memory (on the server or on a separate machine, as desired). An operating system directs the execution of low- level, basic system functions such as memory allocation, file management, and operation of mass storage devices. At a higher level, a control block, implemented as a series of stored instructions, responds to client-originated access requests by retrieving the user-specific profile and applying the one or more rules as described above.
[0104] While there have been described herein what are to be considered exemplary and preferred embodiments of the present invention, other modifications of the invention will become apparent to those skilled in the art from the teachings herein. The particular methods of manufacture and geometries disclosed herein are exemplary in nature and are not to be considered limiting. It is therefore desired to be secured in the appended claims all such modifications as fall within the spirit and scope of the invention. Accordingly, what is desired to be secured by Letters Patent is the invention as defined and differentiated in the following claims.. [0105] What is claimed is:

Claims

CLAIMS 1. A computer- implemented method for generating a test set comprising test stimuli for testing a subject using a computer system comprising logic-based processing circuitry, the method comprising: selecting one or more features from among a plurality of features, wherein a measurable effect on the subject for each feature selected exceeds a predetermined threshold; based on the selected features, generating one or more sample classes; selecting test stimuli from each sample class for presenting to the subject, wherein the selecting step comprises: initially choosing each stimulus from each sample class to present to the subject, wherein the chosen stimuli comprise the test stimuli; and for each sample class that comprises more than one stimulus, subsequently deselecting at least one stimulus from the test stimuli.
2. The method of claim 1, wherein at least one of the deselected test stimuli is deselected randomly.
3. The method of claim 1, further comprising deselecting at least one stimulus associated with a non-selected feature to present to the subject.
4. The method of claim 3, wherein deselecting at least one stimulus associated with a non- selected feature is repeated iteratively.
5. The method of claim 4, wherein iteratively deselecting at least one stimulus associated with a non-selected feature comprises deselecting one less stimulus from a class comprising more than one stimulus.
6. The method of claim 5, wherein iteratively deselecting at least one stimulus associated with a non-selected feature comprises randomly deselecting a stimulus.
7. The method of claim 1, further comprising initially presenting each of the test stimuli prior to selecting from among the plurality of features associated with each of the test stimuli.
8. The method of claim 6, wherein selecting from among the plurality of features comprises computing for a feature a product equal to a value assigned to the feature times a number of stimuli influenced by the feature.
9. The method of claim 1, further comprises determining a measure of significance of each sample class on assessing the hearing capability of the patient.
10. The method of claim 8, wherein the measure of significance of the ith class is equal to
Figure imgf000042_0001
otherwise where nt is the number of stimuli contained in the ith class, andp^ is an empirically determined value based upon an assessment of patient responses to administered test stimuli.
11. The method of claim 1 , further comprises determining the number of times to present different test stimuli to the patient based upon a computed measure of patient hearing weakness, the measure, st, being defined as
s, = k x-
wherein et is a number of test stimuli selected from the ith class and erroneously recognized by the patient, wherein c is a number of test stimuli contained in the ith class, and wherein k is a predetermined constant.
12. The method of claim 1, further comprising the step of presenting the test stimuli to the subject.
13. The method of claim 12, wherein the presenting step further comprises: selecting at least two test stimuli; transmitting a noise stimulus to the subject; transmitting the two test stimuli to the subject; and receiving at least one associated user response to the two test stimuli, wherein the associated user response is based at least in part on the test stimuli.
14. The method of claim 13, further comprising the step of determining a relative weakness of a feature set for the subject, wherein the relative weakness is based at least in part on the noise stimulus and the two test stimuli.
15. The method of claim 13, further comprising the steps of: adjusting a volume of the noise stimuli from the first volume to a second volume; and repeating the noise transmitting step, the test stimuli transmitting step, and the receiving step.
16. The method of claim 15, wherein the adjusting step is based at least in part on the associated user response.
17. The method of claim 15, wherein the first volume is greater than the second volume.
18. The method of claim 13, wherein the associated user response comprises a non- response.
19. The method of claim 13, wherein the at least two test stimuli comprise a phoneme pair.
20. The method of claim 19, wherein the noise stimulus comprises a temporal structure and the test stimuli comprise a temporal structure, and wherein the noise temporal structure substantially matches the test temporal structure.
21. The method of claim 13, wherein the transmitting step and the receiving step together comprise a sequential testing step, wherein the sequential testing step comprises transmitting a first test stimulus, then receiving a first associated user response, then transmitting a second test stimulus, then receiving a second associated user response.
22. The method of claim 13, further comprising the step of determining a threshold noise volume.
23. The method of claim 22, wherein the threshold noise volume comprises the first volume.
24. A computer-implemented method of testing a hearing-impaired patient using a computer system including logic-based processing circuitry, the method comprising the steps of: audibly presenting a plurality of phonemes to the patient, wherein each phoneme is selected from one of a plurality of phoneme sets corresponding to a predetermined feature selected for testing a hearing capability of the patient, wherein the plurality of phonemes comprise test stimuli; based on a response of the patient to each audibly presented phoneme, generating a first assessment of the hearing capability of the patient; and audibly presenting a second plurality of phonemes to the patient and generating a second assessment of the hearing capability of the patient based on patient response, wherein the second plurality of phonemes is selected by deselecting from the test stimuli at least one phoneme from each phoneme set that contains more than one phoneme, wherein at least one of the deselected phonemes is deselected randomly.
25. The method of claim 24, further comprising presenting a third plurality of phonemes, wherein the third plurality of phonemes is selected by deselecting from the test stimuli at least two phonemes from each phoneme set that contains more than one phoneme, wherein at least one of the deselected phonemes is deselected randomly.
26. The method of claim 25, further comprising adjusting at least one operational parameter of a hearing-enhancement device based at least in part on a hearing performance of the subject with respect to different parameter values.
27. A computer-based system for generating a test set comprising test stimuli for testing a subject, the system comprising: at least one processor; an electronic memory having stored therein electronic data representing a plurality of features for testing a particular capability of the subject; a feature-selecting module configured to execute on the at least one processor for selecting one or more features from among the plurality of features, wherein a measurable effect on the subject of each feature selected exceeds a predetermined threshold; and a stimulus- selecting module configured to execute on the at least one processor for generating one or more classes of stimuli based upon the selected feature, and selecting at least one stimulus from each class for presenting to the subject; wherein the stimulus- selecting module is further configured to initially choose each stimulus from each class to present to the subject wherein the chosen stimuli comprise the test stimuli, and subsequently deselecting from the test set at least one stimulus from each class that comprises more than one stimulus, wherein at least one of the deselected test stimuli is deselected randomly by the stimulus-selecting module.
28. The system of claim 27, wherein the stimulus- selecting module is further configured to deselect at least one stimulus associated with a non-selected feature to present to the subject.
29. The system of claim 27, wherein the stimulus-selecting module iteratively deselects at least one stimulus associated with a non-selected feature.
30. The system of claim 28, wherein iteratively deselecting at least one stimulus associated with a non-selected feature comprises deselecting one less stimulus from a class comprising more than one stimulus.
31. The system of claim 29, wherein iteratively deselecting at least one stimulus associated with a non-selected feature comprises randomly deselecting a stimulus.
32. The system of claim 27, wherein the system is configured to initially present each of the predetermined number of stimuli prior to selecting from among the plurality of features associated with each of the predetermined number of stimuli.
33. The system of claim 31, wherein the feature-selecting module is configured to select from among the plurality of features by computing for a feature a product equal to a value assigned to the feature times a number of stimuli influenced by the feature.
34. The system of claim 27, further comprising a significance-determining module for determining a measure of significance of each class of stimuli on assessing the hearing capability of the patient.
35. The system of claim 33, wherein the measure of significance is equal to
Figure imgf000045_0001
otherwise where nt is the number of stimuli contained in the ith class, and /^ is an empirically determined value based upon an assessment of patient responses to administered stimuli.
36. The system of claim 27, further comprising a stimulus-presentment module for determining a number of times to present different stimuli to the patient based upon a computed measure of patient hearing weakness, the measure, S1, defined as
Figure imgf000046_0001
wherein et is a number of stimuli selected from the ith class and erroneously recognized by the patient, wherein c is a number of stimuli contained in the ith class, and where k is a predetermined constant.
37. A computer-based system of testing a hearing-impaired patient with a test set, the system comprising: an audio unit for audibly presenting a plurality of phonemes to the patient, wherein each phoneme is selected from one of a plurality of phoneme sets corresponding to a predetermined feature selected for testing a hearing capability of the patient, wherein the plurality of phonemes comprise the test set; and a testing unit comprising at least one processor for generating a first assessment of the hearing capability of the patient based upon a response of the patient to each audibly presented phoneme; wherein the system audibly presents a second plurality of phonemes to the patient and generates a second assessment of the hearing capability of the patient based upon patient response, wherein the second plurality of phonemes is selected by the testing unit deselecting from the test set at least one phoneme from each phoneme set that contains more than one phoneme, wherein at least one of the deselected phonemes is deselected randomly.
38. A computer-readable medium having embedded therein computer-executable code that, when downloaded and executed by a computer system, causes the computer system to: select one or more features from among a plurality of features, wherein a measurable effect on the subject of each feature selected exceeds a predetermined threshold; based on the selected features, generate one or more classes of stimuli; and select test stimuli from each class for presenting to the subject, wherein the selection of test stimuli comprises initially choosing each stimulus from each class to present to the subject, and subsequently deselecting from the test stimuli at least one stimulus from each class that comprises more than one stimulus, wherein at least one of the deselected phonemes is deselected randomly.
39. A computer-implemented method for generating a test set for testing a subject using a computer system comprising logic-based processing circuitry, the method comprising: selecting one or more features from among a plurality of features, the selecting comprising selecting one or more feature having a measurable effect on the subject, as determined based on a predetermined threshold; based on the selected features, generating one or more classes of stimuli; and selecting stimuli from one or more of said classes for presenting to the subject.
40. A computer-implemented method for generating a test set for testing a subject using a computer system comprising logic-based processing circuitry, the method comprising: selecting one or more features from among a plurality of features, wherein a measurable effect on the subject for each feature selected exceeds a predetermined threshold; based on the selected features, generating one or more classes of stimuli; and selecting stimuli from one or more of said classes for presenting to the subject, wherein selecting comprises choosing a stimulus from a class to present to the subject wherein the chosen stimuli comprises the test set, and subsequently deselecting at least one stimulus from the test set from each class that comprises more than one stimulus.
41. A method of assessing hearing characteristics of a subject, the method comprising: determining a hearing capability of the subject based on responses of the subject to a series of sounds presented to the subject, each sound corresponding to a presence, absence or irrelevance of a predetermined plurality of features; and assigning the subject to one of a predetermined plurality of classes based upon the responses of the subject, each of the plurality of classes being derived from hearing tests performed on a plurality of other subjects.
42. The method of claim 41, further comprising setting one or more parameter values of a hearing-enhancement device based on the class to which the subject is assigned.
43. The method of claim 41, wherein the step of determining hearing capability comprises identifying one or more of the plurality of features as contributing more than other of the plurality of features to a failure of the subject to correctly respond to the presentment of one or more, of the series of sounds.
44. The method of claim 43, wherein a failure to correctly respond to a particular one of the series of sounds defines a feature error with respect to the one or more features corresponding to that particular one of the series of sounds, and further comprising generating a performance measure for the subject based upon a computed mean of feature errors.
45. The method of claim 44, wherein the computed mean of feature errors equals a weighted mean, and further comprising computing the weighted mean, ξ, to be equal to
N
WΛ ξ =
where W1, is a weight assigned to the ith feature of the plurality of features and nt is the number of feature errors with respect to the ith feature.
46. The method of claim 45, wherein the step of assigning the subject to one of the predetermined plurality of classes comprises computing a weighted contribution of each feature, the weighted contribution of a feature quantitatively measuring the contribution that the feature makes to the computed mean of feature errors.
47. The method of claim 46, wherein computing the weighted contribution of a feature comprises computing a value equal to
Contήbution{fι ) = -^ ξ wherein Contήbuύonifj) is the weighted contribution of the ith feature.
48. A system for tuning a hearing-enhancement device, the system comprising: a subject interface for rendering a series of sounds to a subject and for receiving from the subject a response to each of the sounds rendered, each sound corresponding to one or more features belonging to a predetermined plurality of features; and a processing unit communicatively linked, to said subject interface, the processing unit having: a hearing-capability module for determining a hearing capability of the subject based on the received responses of the subject to the series of sounds rendered; a class-assigning module for assigning the subject to one of a predetermined plurality of classes based upon the received responses, each of the plurality of classes being derived from hearing tests performed on a plurality of other subjects; and a tuning module for setting one or more parameters of a hearing-enhancement device based on the class to which the subject is assigned.
49. The system of claim 48, wherein the hearing-capability module is configured to determine hearing capability of the subject by identifying one or more of the plurality of features as contributing more than other of the plurality of features to a failure of the subject to correctly respond to the presentment of one or more of the series of sounds.
50. The system of claim 49, wherein a failure to correctly respond to a particular one of the series of sounds defines a feature error with respect to the one or more features corresponding to that particular one: of the series of sounds, arid wherein the hearing-capability module is further configured to measure a hearing performance of the subject, based on a computed mean of feature errors.
51. The system of claim 50, wherein the computed mean of feature errors equals a weighted mean, and wherein the hearing-capability module is configured to compute the weighted mean, ξ, to be equal to
N
WΛ ξ = N ι=\ where W1 is a weight assigned to the ith feature of the plurality of features and nt is the number of feature errors with respect to the ith feature.
52. The system of claim 51, wherein the class-assigning module is configured to assign the subject to one of the predetermined plurality of classes by computing a weighted contribution of each feature, the weighted contribution of a feature quantitatively measuring the contribution that the feature makes to the computed mean of feature errors.
53. The system of claim 52, wherein the class-assigning module computes the weighted contribution of a feature by determining a value equal to
Contribution(fι ) = ^- £ wherein Contribution^) is the weighted contribution of the ith feature.
54. A computer-readable storage medium in which computer-readable code is embedded, the computer-readable code configured to cause a computing system to perform the following steps when loaded on and executed by the computing system: determine a hearing capability of a subject based on responses; of the. subject to a series of sounds presented to the subject, each sound corresponding to a presence, absence or irrelevance of a predetermined plurality of features; and assign the subject to one of a predetermined plurality of classes based on his responses, each of the plurality of classes being derived from hearing tests performed on a plurality of other subjects.
55. The computer- readable storage medium of claim 54, further comprising; computer- readable code for causing the computing system to set one; or more parameter values of a hearing-enhancement device based upon the class to which the subject is assigned.
56. The computer-readable storage medium of claim 54, wherein the step of determining hearing capability comprises identifying one or more of the plurality of features as contributing more than other of the plurality of features to a failure of the subject to correctly respond to the presentment of one or more of the series of sounds.
57. The computer- readable storage medium of claim 56, wherein a failure to correctly respond to a particular one of the series of sounds identifies a feature error with respect to the one or more features corresponding to that particular one of the series of sounds, and further comprises measuring a hearing performance of the subject based on a computed mean of feature errors.
58. The computer- readable storage medium of claim 57, wherein the computed mean of feature errors equals a weighted mean, and further comprising computing the weighted mean, ξ, to be equal to
Figure imgf000050_0001
where W1 is a weight assigned to the ith feature of the plurality of features and nt is the number of feature errors with respect to the ith feature.
59. The computer-readable storage medium of claim 58, wherein the step of assigning the subject to one of the predetermined plurality of classes comprises computing a weighted contribution of each feature, the weighted contribution of a feature quantitatively measuring the contribution that the feature makes to the computed mean of feature errors.
60. The computer-readable storage medium of claim 59, wherein computing the weighted contribution of a feature comprises computing a value equal to
Contήbution{fι ) = —L-L ξ wherein Contribution(fi) is the weighted contribution of the ith feature.
PCT/US2009/055348 2008-08-29 2009-08-28 System and methods for reducing perceptual device optimization time WO2010025356A2 (en)

Applications Claiming Priority (6)

Application Number Priority Date Filing Date Title
US12/201,598 US9319812B2 (en) 2008-08-29 2008-08-29 System and methods of subject classification based on assessed hearing capabilities
US12/201,492 US9844326B2 (en) 2008-08-29 2008-08-29 System and methods for creating reduced test sets used in assessing subject response to stimuli
US12/201,492 2008-08-29
US12/201,598 2008-08-29
US16445209P 2009-03-29 2009-03-29
US61/164,452 2009-03-29

Publications (2)

Publication Number Publication Date
WO2010025356A2 true WO2010025356A2 (en) 2010-03-04
WO2010025356A3 WO2010025356A3 (en) 2010-10-14

Family

ID=41665198

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2009/055348 WO2010025356A2 (en) 2008-08-29 2009-08-28 System and methods for reducing perceptual device optimization time

Country Status (1)

Country Link
WO (1) WO2010025356A2 (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050027537A1 (en) * 2003-08-01 2005-02-03 Krause Lee S. Speech-based optimization of digital hearing devices
WO2007030402A2 (en) * 2005-08-31 2007-03-15 Tympany, Inc. Stenger screening in automated diagnostic hearing test
US20070286350A1 (en) * 2006-06-02 2007-12-13 University Of Florida Research Foundation, Inc. Speech-based optimization of digital hearing devices
WO2008081446A2 (en) * 2007-01-03 2008-07-10 Hadasit Medical Research Services And Development Ltd. Method, algorithm and device for testing visual acuity

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050027537A1 (en) * 2003-08-01 2005-02-03 Krause Lee S. Speech-based optimization of digital hearing devices
WO2007030402A2 (en) * 2005-08-31 2007-03-15 Tympany, Inc. Stenger screening in automated diagnostic hearing test
US20070286350A1 (en) * 2006-06-02 2007-12-13 University Of Florida Research Foundation, Inc. Speech-based optimization of digital hearing devices
WO2008081446A2 (en) * 2007-01-03 2008-07-10 Hadasit Medical Research Services And Development Ltd. Method, algorithm and device for testing visual acuity

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
DATABASE EMBASE [Online] ELSEVIER SCIENCE PUBLISHERS, AMSTERDAM, NL; 1974, NAKAMURA K ET AL: "Auditory perception of duration of speech (and non speech stimuli in hard of hearing subjects) (Japanese)" XP002575735 Database accession no. EMB-1975184281 *

Also Published As

Publication number Publication date
WO2010025356A3 (en) 2010-10-14

Similar Documents

Publication Publication Date Title
CN110072434B (en) Use of acoustic biomarkers to assist hearing device use
US7206416B2 (en) Speech-based optimization of digital hearing devices
US9666181B2 (en) Systems and methods for tuning automatic speech recognition systems
US9319812B2 (en) System and methods of subject classification based on assessed hearing capabilities
JP5129954B2 (en) Cochlear implant system with MAP optimization using genetic algorithm
US20070286350A1 (en) Speech-based optimization of digital hearing devices
US9345881B2 (en) Genetic algorithm based auditory training
Lawson et al. Speech audiometry
US20220386902A1 (en) Scoring speech audiometry
US20220036878A1 (en) Speech assessment using data from ear-wearable devices
US9844326B2 (en) System and methods for creating reduced test sets used in assessing subject response to stimuli
Souza et al. Contributions to speech-cue weighting in older adults with impaired hearing
Skinner et al. Identification of speech by cochlear implant recipients with the multipeak (MPEAK) and spectral peak (SPEAK) speech coding strategies II. Consonants
Scherer et al. Automatic verification of emotionally stressed speakers: The problem of individual differences
AU2009279764A1 (en) Automatic performance optimization for perceptual devices
Müller et al. Comparison of the effects of two cochlear implant fine structure coding strategies on speech perception
WO2010025356A2 (en) System and methods for reducing perceptual device optimization time
CN106859660B (en) A kind of design method and application of closed type speech audiometry vocabulary
Sedigh Application of polyscale methods for speaker verification
Najnin et al. Identifying hearing loss from learned speech kernels.
Arnoult Whole Word vs Phoneme Error Scoring for Audiological Word Recognition Testing
Trevino Techniques for understanding hearing-impaired perception of consonant cues
CN116269360A (en) Remote test method, system, terminal and storage medium for hearing fine capability
Van Zyl Objective determination of vowel intelligibility of a cochlear implant model
Jürgens A microscopic model of speech recognition for listeners with normal and impaired hearing

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 09752541

Country of ref document: EP

Kind code of ref document: A2

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 09752541

Country of ref document: EP

Kind code of ref document: A2