RU2446741C1 - Method of estimating disturbances of aural perception of speech signals - Google Patents

Abstract

FIELD: medicine.

SUBSTANCE: invention relates to field of medicine, namely to otholaryngology. Set of test acoustic speech signals with individual level of loudness is presented and probability of correctly recognised words is calculated. Test speech signals are preliminarily subjected to non-linear frequency distortions of such level which is determined at the stage of preliminary examination of group of people with otologically normal and diminished hearing. For this purpose sum of probabilities of unrecognised test speech stimuli in group of people with otologically normal hearing and correctly recognised test speech stimuli in group of people with diminished hearing is calculated at different levels of nonlinear frequency distortions. Level of nonlinear frequency distortions at which sum of said probabilities is minimal is selected and individual level of loudness is set above the level of threshold of examined person's audibility. If probability of correctly recognised words is lower than probability of correct recognition of set of test acoustic speech signals for examined people with otologically normal hearing, disturbance of auditory perception of speech signals is diagnosed.

EFFECT: method makes it possible to increase reliability of estimation of disturbances of speech signal auditory perception.

5 dwg, 3 ex

Description

The invention relates to medicine, more specifically to otorhinolaryngology, and can be used for early diagnosis of sensorineural hearing loss and determining the state of hearing during occupational selection.

From the prior art, methods are known for assessing violations of auditory perception of speech signals based on changes in volume, phonemic components, signal-to-noise ratio of speech signals presented to a test subject, as well as methods based on assessing interhemispheric coherence of an electroencephalogram (EEG).

A method for assessing violations of auditory perception of speech signals based on a change in volume is implemented in a method for determining the level of speech intelligibility, which consists in calculating the percentage of correctly recognized words at different volume levels of the presented sets of speech signals, and hearing impairment is diagnosed by reducing the proportion of correctly recognized words (Sagalovich BM Methods of hearing research in clinical audiology. In the book: Hearing loss (edited by Preobrazhensky NA). M: Medicine. 1978. S.10-167), (Blagoveshchenskaya NS Oto evrologicheskie symptoms and syndromes -.. M .: Meditsina, 1990. S.24-47), [2], (US 7143031, IPC G10L 11/00, 704/224 NKI, 28.11.2006) [3]. Inspection with this method takes a long time, which is associated with the determination of at least three values of volume levels. In addition, the method is effective with a high degree of hearing loss and does not allow to detect early hearing impairment.

A method for assessing violations of auditory perception of speech signals based on a change in the phonemic components of test signals is implemented in a method for early detection of dyslexia and dysgraphia and a risk group for a disease in children, in which words are presented in the form of a dichotic pair, in one of which the phoneme is replaced by a non-speech signal . When identifying no more than 10% of the correct answers, dyslexia or dysgraphia or a predisposition to their occurrence is diagnosed (RU 2138987 C1, 6 IPC АВВ 5/16, publ. 10.10.1999). The method is aimed at identifying impaired speech functions, caused primarily by dysfunction of cognitive cortical processes, and therefore does not allow to differentiate hearing loss from cortical impairment, which makes it difficult to understand speech, which is necessary for choosing a treatment method.

A method for assessing violations of auditory perception of speech signals, based on a change in the signal-to-noise ratio, is implemented in a method for objectively evaluating deafness and a high degree of hearing loss using computer phonography, which compares the phonogram amplitude of an arbitrary text read by the subject in silence and against the background of noise an ear. If the amplitude of the phonogram of the text being read against the background of the noise has not changed, then a conclusion is made about deafness in the studied ear or about a high degree of hearing loss. (RU 2234245 C2, 6 IPK А61В 5/12, publ. 08/20/2004). However, the known method is effective only with a high degree of hearing loss and does not allow to detect early hearing impairment.

A method based on the assessment of interhemispheric coherence of EEG is used for early diagnosis of occupational sensorineural hearing loss, in which average values of interhemispheric coherence of EEG activity are calculated (RU 2394481, 8 IPC А61В 5/0476, published on July 20, 2010). The initial development of hearing loss is determined if the value of the interhemispheric coherence of the EEG differs from the normative values for a group of people with the corresponding experience. The presence of high individual and interindividual variability of interhemispheric EEG coherence can lead to an erroneous diagnosis of sensorineural hearing loss.

Closest to the technical nature of the claimed invention is a method for evaluating noise immunity of the perception of speech signals, in which they evaluate the ability to correctly identify words presented against the background of masking noise at a certain ratio of the intensities of the test signal and interference. (RU 2147833, 6 IPC А61В 5/12, publ. 04/27/2000), taken as a prototype.

At the preliminary stage of the examination, individually for each individual subject the effective level of presentation of several sets of words is determined, at which the greatest probability of the correct recognition of the words is recorded. Then, test speech signals are presented together with acoustic noise, while increasing the signal-to-noise ratio from zero to a value at which the probability of correct recognition of test signals reaches a value 20% lower than the initial one, taking into account the effective volume level, the signal-to-noise ratio is determined and at its value less than 7 dB is taken as noise immunity for patients with a lack of sensorineural hearing loss, and with a value equal to or more than -7 dB for patients with signs of sensorineural hearing loss tee.

A significant disadvantage of the prototype method is a lengthy multi-stage examination procedure. The preliminary examination stage, at which the effective volume level is selected, requires the presentation of several sets of words, each of which includes 50 words. The main examination procedure also involves the use of several similar-sized sets of words. Thus, the total examination time for one person is several tens of minutes, which is unacceptable for mass screening of the population (screening).

The prototype method does not allow to identify those auditory disorders that are caused by a decrease in auditory sensitivity in certain frequency ranges, which leads to impaired auditory perception of speech signals. In this case, the signal-to-noise ratio is determined for an average speech spectrum with a maximum in the region of 400-700 Hz and a decrease towards high frequencies with a slope of 6 dB / oct., Which is implemented in a device for measuring noise immunity of speech perception by a person of the same authors and applicant (RU 2148391, 6 IPC А61В 5/12, publ. 10.05.2000). Thus, the averaged interference spectrum does not correspond to the real spectrum of a particular test acoustic speech signal and the interference does not equally mask the individual frequency components of the test signal used. This leads to a decrease in the reliability of the assessment of impaired auditory perception, which can be caused by a decrease in auditory sensitivity in certain narrow frequency ranges.

The technical result of the present invention is to increase the reliability of the assessment of impaired auditory perception of speech signals while reducing auditory sensitivity in individual narrow frequency regions due to the distribution of frequency distortions of the speech signal over the entire spectrum of the speech signal, as well as reducing the time of examination of the patient by reducing the number of test speech signals .

The specified technical result is achieved in that the method for assessing violations of the auditory perception of speech signals includes presenting the subject with a set of test acoustic speech signals with an individual volume level and calculating the probability of correctly recognized words.

According to the invention, test speech signals are preliminarily subjected to nonlinear frequency distortion of a level determined at the stage of preliminary examination of groups of people with otologically normal and hearing impairment, for which the sum of the probabilities of unrecognized test speech stimuli in a group of people with otologically normal hearing and correctly recognized test speech is calculated stimuli in the group of persons with hearing loss at various levels of nonlinear frequency distortion and choose the level of nonlinear frequencies distortion-frequency at which the sum of these probabilities is minimal, and the individual volume level is set higher than the auditory threshold of the subject, and the probability of correctly recognized words less probability of correct recognition of a set of test of acoustic speech signals for subjects with normal hearing otologic diagnose a violation of auditory perception of speech signals.

In the particular case of the method:

- the individual volume level is set to 30-40 dB above the threshold of audibility of the subject;

- the level of nonlinear frequency distortion is 1/12 octave;

- the probability of correct recognition of a set of test acoustic speech signals for subjects with otologically normal hearing is 0.9.

The introduction of nonlinear frequency distortions, which are distributed throughout the spectrum of the test speech signal, allows us to detect hearing loss for all components of the spectrum, which leads to an increase in the reliability of the assessment of auditory perception disorders, which can be caused by a decrease in auditory sensitivity in individual narrow frequency ranges.

Using a preliminary examination of groups of people with otologically normal and hearing impaired to form a set of test speech signals allows for an individual examination, in contrast to the prototype, in which the subject is presented with several sets of test speech signals, only one set of test speech signals is presented, which leads to a decrease several times the time of an individual examination.

An increase in the volume of the test speech signal by 30–40 dB above the hearing threshold of the subject provides 100% speech intelligibility in individuals with an otologically normal hearing.

The level of nonlinear frequency distortion of test speech signals of 1/12 octave provides the minimum value of the sum of the probabilities of false positive and false negative errors in the diagnosis of auditory disturbances in speech signals.

The probability of correct recognition of a set of test acoustic speech signals for subjects with otologically normal hearing, equal to 0.9, was obtained experimentally and ensures the separation of people with normal and reduced hearing with a minimum sum of the probabilities of false-positive and false-negative errors in the diagnosis of auditory perception of speech signals.

The operation of the method is illustrated by drawings.

Figure 1 shows the functional block diagram of the implementation of the proposed method for assessing violations of auditory perception of speech signals.

Figure 2 shows the amplitude spectrum of the test speech signal.

Figure 3 shows the amplitude spectrum of the test speech signal with non-linear frequency distortion.

Figure 4 shows histograms of the relative frequencies of correctly recognized words by the examined with otologically normal hearing and sensorineural hearing loss from the level of nonlinear frequency distortion, where the dark bars correspond to the relative frequencies of correctly recognized words by the examined with otologically normal hearing, the shaded bars correspond to the relative frequencies of correctly recognized words by the examined with sensorineural hearing loss.

Figure 5 shows the dependence of the accumulated relative frequencies of correctly recognized words by the examined with otologically normal hearing and sensorineural hearing loss on the level of nonlinear frequency distortions, where curve 1, indicated by the dashed line, corresponds to the relative frequencies of correctly recognized words, examined by the otologically normal hearing, and curve 2, denoted by a solid line, corresponds to the relative frequencies of correctly recognized words being examined with sensorineural hearing loss.

Figure 6 shows the dependence of the probability of false-negative errors in the diagnosis of disorders of auditory perception of speech signals, corresponding to the probability of correctly recognized words in people with impaired auditory perception of speech signals, the probability of false-positive errors in the diagnosis of violations of auditory perception of speech signals, corresponding to the addition to a unit of probability of recognized words in people with otologically normal hearing, and their sums from the level of nonlinear frequency distortion, where curve 1 corresponds to the number of false negative errors and false positive errors in the diagnosis of auditory perception of speech signals, curve 2 corresponds to the probability of false positive errors in the diagnosis of auditory perception of speech signals, and curve 3 corresponds to the sum of the probabilities of false negative errors and false positive errors in the diagnosis of auditory perception of speech signals.

The block diagram of the proposed method for assessing violations of auditory perception of speech signals (Fig. 1) is implemented programmatically on a personal computer of type PC and contains a block for storing test signals 1, a block for sampling test signals 2, an amplification block for test signals 3, headphones 4, and the patient ) 5, operator (doctor) 6, response registration unit 7, probability determination unit for correct answers 8, decision making unit 9, control unit 10, examination results display unit 11. Audi was used as head phones 4 Metric headphones TDH-39P firm «Telephonics». The subject 5 is located in a sitting position with the head phones 4 on, which are connected to the test signal amplification unit 3.

A pre-formed set of test acoustic speech signals is introduced into the test signal storage unit 1.

The procedure for forming a set of test acoustic speech signals was as follows. As the initial test acoustic speech signals, 50 words were used, taken from articulation tables of phonetically balanced polysyllabic words of Greenberg-Zinder (Grinberg G.I., Zinder L.R. Word tables for speech audiometry in clinical practice. // Proceedings of the Leningrad Scientific Research Institute for Diseases ear, throat, nose, nose. - L. - 1957 - v. 2. - P.45-47), recorded by a professional male speaker on magnetic media. These analog signals were converted to digital signals and normalized by the level of the rms power of the frequency spectrum, then subjected to degree nonlinear frequency distortion within one octave by the well-known Pitch Band method of uniform frequency shift across the entire spectrum of the signal using the Sony Sound Forge 7.0 sound editor from Sonic Foundary Inc. For each source signal, a set of 24 gradations of distortion was obtained. Thus obtained 50 × 24 = 1200 test acoustic speech signals were stored in the computer memory and then presented for recognition to a group of 96 persons aged 20 to 25 years with otologically normal hearing and a group of 48 persons of the same age with a diagnosis of sensorineural hearing loss. For each of 50 sets of 24 gradations of distortions of the source signal, the maximum distortion level was determined at which the subjects correctly recognized the word. The signals were presented at an above-threshold volume level, which exceeded the individual detection threshold by 30 dB in sequence from the maximum distortion level (24 level), which corresponds to a frequency change of one octave to zero level, which corresponds to the absence of distortion. According to the results of the survey using statistical calculations (Johnson N., Lyon F. Statistics and experimental design in engineering and science. Data processing methods. Mir Publishing House, Moscow, 1980, p. 26-29.), The relative frequencies of correctly recognized words (figure 4) and built storage frequency polygons (figure 5). To minimize false-negative and false-positive errors in the diagnosis of auditory perception of speech signals, we calculated the sum of the probability of correct answers of people with sensorineural hearing loss at a given level of distortion (the probability of false negative errors in the diagnosis of auditory perception of speech signals) and the probability of the absence of correct answers in people with normal hearing (the probability of false positive diagnostic errors violations of auditory perception of speech signals) (Fig.6, curve 3). From the graph in Fig.6 it follows that the minimum sum of errors (curve 3) corresponds to the presentation of signals with level 2 of nonlinear frequency distortion.

As follows from the graph in Fig. 5 (curve 1 for a group of subjects with an otologically normal hearing), the probability of correctly recognized words is 0.9 at the level of non-linear frequency distortions 2.

From a set of 1200 test speech signals, signals with a found distortion level are selected and a final set of 50 test speech signals is formed, which are input into the test signal storage unit and used in assessing auditory disturbances in speech signals.

At the command of the control unit 10, which enters the block of sampling test signals 2, the signal of a given number is read from the memory block of the test signals 1 and enters the block amplification of test signals 3, where it is amplified in accordance with the command received from the control unit 10.

The intensity of the test speech signals in the amplification unit of the test signals 3 is set 30 dB above the threshold for their detection. The detection threshold is determined before the diagnostic examination procedure by the known method of constant stimuli (Sagalovich BM Methods of hearing research in clinical audiology. In the book: Hearing loss (edited by Preobrazhensky NA). M: Medicine. 1978. S. 61). The required gain level is stored in the control unit 10 and is used to set the gain in the amplification unit of test signals 3, from the output of which the test speech signal is sent to the headphones 4. Surveyed 5 in response to the heard signal calls the operator 6 the word that he heard or reports about inability to recognize the test signal. Operator 6 compares the word named by examinee 5 with the correct answer option stored in the memorizing unit 1 and registers the correct answer by pressing the key using the response recording unit 7. After registering the answer of the examined 5 in the probability determination block of correct answers 8, the probability of the correct answers of the examined 5 is calculated , and in the control unit 10, a command is generated to supply the next test signal. When the number of responses of examinee 5 is equal to the number of test signals specified in the control unit 10, a command is sent from control unit 10 to the unit for determining the probability of correct answers 8 to transmit the probability of correct answers of the examinee to decision block 9. The probability of correct answers of examinee 5 and the result of decision making the presence of violations of the auditory perception of speech signals is displayed in the display unit 11. The auditory disturbances in speech signals are diagnosed when The probability of correct recognition of words from the presented set is less than 0.9.

Example 1

Surveyed K.S., 21 years old. Speech audiometry performed using the MA-31 audiometer showed 100% speech intelligibility at a volume level of 20 dB above the hearing threshold, which is normal. The examination time was 11 minutes. To test the proposed method through headphones, the test subject was presented with a set of 50 words with non-linear frequency distortions. The probability of correct word recognition was 0.92 (speech intelligibility 92%), which confirms the normal state of hearing. The examination time by the claimed method was 3 minutes.

Example 2

Examined S.O., 52 years old. Speech audiometry performed using the MA-31 audiometer showed 100% speech intelligibility at a volume level of 22 dB above the hearing threshold, which is normal. The examination time was 14 minutes. Speech intelligibility, determined by the claimed method, was 82%, which corresponds to impaired perception of speech signals. Additionally, tonal audiometry revealed a significant decrease in hearing in the right ear for a frequency of 6000 Hz by 45 dB, which indicates the initial stages of the development of sensorineural hearing loss. The examination time by the claimed method was 3 minutes.

Example 3

Examined D.V., 65 years old. Speech audiometry performed using the MA-31 audiometer showed 100% speech intelligibility at a volume level of 25 dB above the hearing threshold, which is normal. The examination time was 12 minutes. Speech intelligibility according to the claimed method was 74%, which corresponds to hearing impairment. Additionally, tonal audiometry revealed hearing loss in the right ear for a frequency of 3000 Hz by 40 dB and in the left ear for a frequency of 4000 Hz by 50 dB, which confirmed the development of sensorineural hearing loss in the subject. The examination time by the claimed method was 4 minutes.

The inventive method was successfully tested on a group of 126 students, 5 people had a violation of auditory perception of speech signals, confirmed by subsequent clinical examinations.

Information sources

1. Blagoveshchenskaya N.S. Otoneurological symptoms and syndromes. - M.: Medicine. 1990. S. 270-271.

2. Sagalovich B.M. Methods of hearing research in clinical audiology. In the book: Hearing loss (edited by Preobrazhensky N.A.). M .: Medicine. 1978. S. 61-69.

3. RU 2138987 C1, 10.10.1999, 6 A61B 5/16.

4. RU 2234245 C2, 20-08-2004, A61B 5/12.

5. RU 2394481 C1, 07.20.2010, A61B, 5/0476.

6. US 7143031, IPC G10L 11/00, NCI 704/224, 11.28.2006.

7. RU 2147833 C1, 04/27/2000, A61B 5/12 - prototype.

8. RU 2148391 C1, 05/10/2000, A61B 5/12.

9. Greenberg G.I., Zinder L.R. Word tables for speech audiometry in clinical practice. // Proceedings of the Leningrad Scientific Research Institute on diseases of the ear, throat, nose and nose. - L. - 1957 - v. 2. - Pages 45-47.

10. Johnson N., Lyon F. Statistics and experimental design in engineering and science. Data processing methods. Publishing house "Mir", Moscow, 1980, S. 26-29.

Claims (4)

1. A method for assessing violations of auditory perception of speech signals, comprising presenting the subject with a set of test acoustic speech signals with an individual volume level and calculating the probability of correctly recognized words, characterized in that the test speech signals are preliminarily subjected to nonlinear frequency distortions of the level determined at the preliminary examination stage groups of individuals with otologically normal and hearing impairment, for which the sum of the probabilities of unrecognized speech stimuli in the group of people with otologically normal hearing and correctly recognized test speech stimuli in the group of people with hearing loss at different levels of nonlinear frequency distortion, and choose the level of nonlinear frequency distortion at which the sum of these probabilities is minimal and the individual volume level is set above the threshold the audibility of the subject, and when the probability of correctly recognized words is less than the probability of correct recognition of a set of test acoustic speech signals for those with otologically normal hearing are diagnosed with impaired auditory perception of speech signals. 2. The method according to claim 1, characterized in that the individual volume level is set to 30-40 dB above the audibility threshold of the subject. 3. The method according to claim 1, characterized in that the level of non-linear frequency distortion is 1/12 octave. 4. The method according to claim 1, characterized in that the probability of correct recognition of a set of test acoustic speech signals for subjects with otologically normal hearing is 0.9.

Images