EP2540099A1

EP2540099A1 - Method for training speech recognition, and training device

Info

Publication number: EP2540099A1
Application number: EP10775754A
Authority: EP
Inventors: Maja Serman; Martina Bellanova
Original assignee: Siemens Medical Instruments Pte Ltd
Current assignee: Sivantos Pte Ltd
Priority date: 2010-02-24
Filing date: 2010-10-21
Publication date: 2013-01-02
Also published as: US20110207094A1; AU2010347009B2; WO2011103934A1; US20130209970A1; AU2010347009A1

Abstract

The aim of the invention is to improve speech recognition in wearers of hearing aids and other hearing devices. This is achieved in that a method is provided for training the speech recognition of a person wearing a hearing device. A first speech element is acoustically presented, and said element is identified by the person wearing the hearing device. Subsequently, the acoustic presentation of the presented speech element is automatically changed and the aforementioned steps are repeated (S1 to S4) with the changed presentation until a specified maximum number of repetitions is reached if the identification is incorrect. Otherwise, a second speech element is acoustically presented if the identification of the first speech element is correct or if the number of incorrect identifications of the first speech element is greater than the maximum number of repetitions. In this manner, each of a plurality of speech elements can be trained in multiple stages.

Description

description

Method for training the speech comprehension and training ^¬ device

The present invention relates to a method for Trainie ^¬ ren of a person's speech understanding that carries a Hörvorrich ^¬ tung, person through the bearing by acoustically presenting a speech inventory ^¬ part and identifying the voice notification acoustically presented to the hearing apparatus standing piece. In addition, the present invention relates to a Vorrich ^¬ device for automatically training the speech understanding of a person wearing a hearing, with a playback ^¬ device for the acoustic presentation of a first Sprachbe- part and an interface device for inputting an identifier for identifying the acoustically presented language component by the person wearing the hearing device. A hearing device here means any sound-emitting device which can be worn in or on the ear, in particular a hearing device, a headset, headphones, loudspeakers and the like.

Hearing aids are portable hearing aids that are used to care for the hearing impaired. In order to meet the numerous individual needs, different types of hearing aids such as behind-the-ear hearing aids (BTE), hearing aid with external receiver (RIC: receiver in the canal) and in-the-ear hearing aids (IDO), e.g. Concha hearing aids or canal hearing aids (ITE, CIC). The hearing aids listed by way of example are worn on the outer ear or in the ear canal. In addition, bone conduction hearing aids, implantable or vibrotactile hearing aids are also available on the market. The stimulation of the damaged hearing takes place either mechanically or electrically.

Hearing aids have in principle as essential components an input transducer, an amplifier and an output transducer. The input transducer is usually a sound catcher, z. As a microphone, and / or an electromagnetic receiver, for. B. an induction coil. The output transducer is usually used as an electroacoustic transducer, z. As miniature speaker, or as an electromechanical transducer, z. B. bone conduction, implemented. The amplifier is usually integrated in a signal processing unit. This basic structure is shown in FIG. 1 using the example of a behind-the-ear hearing device. In a hearing aid housing 1 for carrying behind the ear, one or more microphones 2 for receiving the sound from the environment are installed. A signal ^¬ processing unit 3, which is also integrated into the hearing aid housing 1 processes the microphone signals and ver ^¬ strengthens them. The output signal of the signal processing unit 3 is transmitted to a loudspeaker or earpiece 4, which outputs an acoustic signal. The sound is optionally, for the wearer's eardrum übertra ^¬ gene via a sound tube which is fixed with a otoplasty in the auditory canal. The energy supply of the hearing aid and in particular the signal processing unit 3 by a likewise integrated into the hearing aid housing 1 battery 5.

The speech intelligibility plays an out ^¬ dominant role in hearing aids. When transmitting the sound through a hearing aid, the sound is changed. In particular takes place beispielswei- se a frequency compression, like a dynamic compression (compression of the input level range to the Ausgangspegelbe ^¬ rich), a noise reduction and the like. In all these processes, voice signals are also changed, which ultimately makes them sound different. In addition, the speech intelligibility of a subject worsens due to his hearing loss. This can be demonstrated, for example, by speech audiograms.

De Filippo and Scott, JASA 1978, make known a so-called "connected discourse test." This test is the most widely used, non-PC-based, language comprehension training. Training is done on a word-by-word basis Attention and possibly an intervention of the trainer or tester. The differing ^¬ chen levels depend on intentional and accidental factors, which result from the tester, namely voice type, volume changes and the like. The test is very tiresome for the test person and the tester and is therefore limited in practice to five to ten minutes.

The object of the present invention is to improve speech understanding through targeted training and to perform this training as automated as possible.

According to the invention, this object is achieved by a method for automatically training the speech comprehension of a person wearing a hearing device

a) acoustic presentation of a first voice component and b) identification of the acoustically presented voice component by the person wearing the hearing device, and c) automatic modification of the acoustic performance of the person

presented language component and repeating the

Steps a) and b) with the change in performance when the identification is incorrect, repeatedly until a pre give ^¬ ne maximum iterative number is reached, and

d) acoustically presenting a second voice component if the identification of the first voice component is correct or the number of false identifications of the first voice component is greater than the maximum repetition number by one.

In addition, the invention provides a device for automatically training the speech comprehension of a person wearing a hearing device

a) a reproducing device for acoustically presenting a first language component and

b) interface means for inputting an identifier (e.g., acoustic response or manual input) for

Identifying the acoustically presented language component by the person wearing the hearing device, as well as c) control means which controls the reproducing device and the interface device that akusti ^¬ specific performance of the speech component is automatically changed and the steps a) and b) are repeated with the change in performance when the identification is incorrect, repeatedly until a predetermined maximum is re ^¬ holungsanzahl reached, and a second speech inventory ^¬ part is presented, when identifying the first language constituent correct or the number of false identifications of the first speech component is one greater than the maximum number of repetitions.

In an advantageous manner, a change in the performance ensues when the same voice component is reproduced again acoustically. This leads to an improved

Training effect. In particular, this corresponds to the natural situation that the same words are presented in many different ways to the hearing. Conveniently, logatoms or words are used to train speech understanding. A logatom is one out

Speech sounds composed of verbs such as "atta", "assa" and "ascha." Each log atom can consist of several phonemes, where a phoneme represents an abstract class of all sounds that have the same meaning-discriminating function in a spoken language.

With the logatomes an efficient training can be carried out on a very low level of complexity. This can also be automated more easily, with the automatic

Feedback of the recognition or non-recognition of a dargebo ^¬ tenen test word or -logatoms reinforces the learning effect.

In one embodiment, a number of Sprachbestandtei- len is predetermined, and the steps a) to d) be repeated as often as ^¬ until all the voice components are at least once Darge ^¬ offered. That way, in a training session a fixed set of logatomes or words are trained.

In particular, in repetition, the voice component may be presented with greater emphasis than the first performance. In one variant, in repeating the speech component may be presented to the previous performance with a different voice or background noise. In this way, a hearing aid user can be prepared for the different natural situations, for example when its call partner has been said difference ^¬ Lich articulate or if on the one hand a male voice and a female voice on the other hand is presented to him.

Furthermore, at the beginning of the procedure, the language component may be a logatom and at its last repetition a word in which the logatom is integrated. When the Loga ^¬ tom is in a word, understanding the Logatoms is simplified because it is perceived in context.

Specifically, identifying the voice component rendered altered by the hearing device may be performed by the person having a graphical user interface. The person or the subject must then select "multiple choice test" a writing presented several variants only in accordance with a. This what is understood can be differentiated exact circumstances. Be at a fal ^¬ rule identification In another embodiment of the proffered and of the person specified language component acoustically and / or optically and ^¬ dergegeben. Due to the acoustic playback of both variants the person immediately gets an audible or auditory comparison of the heard and the reproduced speech component. This learning is simplified. This can even through the visual reproduction of the Both variants are supported. In a likewise preferred embodiment, the person always receives the voice component at a constant volume from the hearing device. This turns off a variable, namely the volume, during training. Thus ^¬ with the person does not become influenced during speech understanding that the spoken word is presented at different volumes. Conveniently, all process parameters are determined in advance by a trainer and sent to the person to be trained by the trainer. Thus, the training can be done comfortably for a hörge ^¬ damaged person. Furthermore, be ^¬ this means that the training can take place essentially without intervention of a tester. This in turn has the advantage that the tester the result impartially and whether ^¬ can evaluate objectively compared to other results.

The present invention will now be explained in more detail with reference to the accompanying drawings, in which:

1 shows the basic structure of a hearing aid according to the

State of the art; 2 shows a schematic diagram of a training sequence; and

3 shows a schematic diagram for determining a erfindungsge ^¬ MAESSING training process. The embodiments described in more detail below represent preferred embodiments of the present invention.

In FIG. 2, the course of a possible variant for training speech understanding is shown symbolically. One person undergoes training or testing. It receives voice components from a voice output device 11 (eg loudspeakers in a room or headphones), in particular logos. me 12, presented. For example, such a logatom is spoken by a man or a woman with one or the other emphasis. The logatom 12 is picked up by the hearing device or the hearing aid 13, which carries the person 10, and amplified specifically for the hearing impairment of the person. In this case, a corresponding frequency compression, dynamic compression, noise reduction or the like. The hearing aid 13 acoustically outputs an altered logatom 14. This altered logatom 14 reaches the hearing of the person 10 as a modified acoustic performance.

The hearing aid wearer, ie the person 10, attempts to understand the acoustically altered logatom 14 presented to him verbally. He is a graphical user interface 15 available. For example, various solutions are presented to the person 10 on this graphical user interface 15. Several logatoms will show up as Antwortal ^¬ ternatives writing. The choice of answer alternatives may be based on phonetic similarity or other criteria as appropriate. She then chooses the one of the recorded logatoms that she thinks she has understood. The result of the selection by the person 10 may, for example be held in a Verwechs ^¬ lung matrix sixteenth In it, the presented logatomes are represented opposite to the recognized logatomes. As the dashed arrow 17 indicates in Figure 2, the test can be repeated unchanged or changed. In particular ^¬ sondere other logatoms or the same logatoms can be presented in a different way in the repetition.

The intelligibility of speech training is, as indicated above, preferably implemented on a computer with graphical Benut ^¬ cut point. It can be developed for example in ei ^¬ ner MATLAB environment.

The implemented test or training method can be implemented in n (preferably four) training levels with acoustic feedback (confirmation or notification of an error) be. In a first stage of training the volunteers or the person is a Logatom or word as sound as dargebo ^¬ th. The person is asked to choose one answer among five eg visually presented alternatives. If the person gives the correct answer, the sound example of how ^¬ is derholt and a "correct" message is shown as feedback. The person may be the correct answer be repeated if it has, for example, only guess at the answer. In case of right If the person makes a mistake, they will get an acoustic feedback with a comparison of the choice and the correct answer (eg "You have assa ' ^¬ geant ^¬ say and we played 'affa'. "). This feedback can also be repeated as often as desired. After the mistake, the person goes to the second training level.

Because of the error, the person has to go through the second training stage, where the same sound example as in the previous stage is presented. However, the performance is in a different difficulty mode. For example, understanding is facilitated by speech reproduction with clear speech or overemphasis. For training purposes, the emphasis can also be reduced. After playing the sound example, the person has to choose one of eg five alternatives as his answer. If the person chooses the correct answer, the sound example (Logatom) is repeated, and it will appear as a return ^¬ coupling or issued a "correct" message. The person may repeat as desired the correct answer so often. From here, the person goes to the next sound example as in the first training stage. However, if the person makes a mistake, she gets, also as in the first Trainingsstu ^¬ fe, acoustic feedback with a comparison of their choice and the correct answer. these feedback can repeated as many times as desired, due to the error the person has to go to a third training level etc. In the present embodiment, a total of n training levels are provided. If the person does not understand the sound example in the nth training step (nth repetition) (nth false identification), it is registered in a test protocol. At the end of the training, all sound examples that were not understood in all n training levels can be tested or trained again in n training levels. The training process (training mode) can be performed with increasing, decreasing or constant difficulty level. Different difficulty ^modes are, for example, a female voice, a male voice, a clear voice of a male voice, a clear voice of a female voice, an additional word description, noise reduction, etc.

It can be given a fixed training set with adjustable ^{bar ¬} number of sound examples and adjustable number of answer alternatives per sound example. In addition, the test or training can be performed at rest or with different background sounds (static or modulated, depending on the purpose of the test).

It will be explained with reference to FIG. 3 how a training sequence can be determined, for example, by an acoustician. With the aid of a user interface 20, the acoustician can specify ^different parameters for the training procedure. To ^¬ closest he chooses, for example, 21. This example, a VCV or CVC-type or both (vowel-consonant-vowel or consonant-vowel-consonant) the Phonemtyp. He can also specify a particular vowel 22 for the selected phoneme type.

The training consists of four stages Sl to S4, as in the previous example. At each stage, the acoustician has the opportunity to adjust or tune the difficulty of the performance (23). In this background ^¬ noise can be simulated in different listening situations, for example. Furthermore, the acoustician, for example, the voice source

Set 24 for each training level Sl to S4. Beispielswei ^¬ se can choose a male or female language. However, the votes of different men or different women may also be taken. Optionally, the emphasis can also be varied. Advantageously, one of the parameters 23, 24 is varied at least from one to the next learning ^¬ stage Sl to S4. In ei ^¬ nem concrete example of the difficulty level 23 remains the same in all stages, but it is as a source 24 in step Sl a female voice to the presentation of an Logatoms, in step S2, a male voice for presenting a Logatoms, in step S3 a clear male voice for presenting a Logatoms and in step S4 a word containing the log atom presented.

Finally, the acoustician or coach's feedback

Configure 25 for the exerciser. For this acti ^¬ fourth the acoustician for example, which are the remaining, yet to be trained logatoms or words Toggle display. In addition, he can set whether the feedback 25 should be purely visual or audible. He can also specify whether correct answers are marked in the overall evaluation. Other process parameters can also be defined in this way.

Below are some technical details darge ^¬ represents, with which the test can be equipped. In a preferred embodiment, the test is not adaptive, but performed at a constant volume level. This allows the person to focus on learning the processed speech signal without having to adjust or learn the volume level as well. The language has namely acoustic characteristics (spectral changes) that are to learn inde ^¬ pending on the volume changes (which must also be learned). The learning effect increases when the two aspects are separated. With regard to the training levels, repetition is already a way of learning. The feedback is automatic after an error and the person can repeat the language example. In addition to the repetition itself, there are n successive stages of learning where you can choose whether you want a simple repetition or a change in the difficulty mode of the stimulus. When Su ^¬ alteration of difficulty mode from difficult to easy for the same sound as learning is facilitated. It has been found that changing the voice of the speaker increases the learning effect. In addition, the learning effect can also be increased in that the Schallbei ^¬ game is embedded in a context (context of a sentence). All these effects can be combined to increase or decrease the learning difficulty.

All test options are determined in a further embodiment in advance independently of the test procedure and stored in a settings file. Thus, the test can be carried out, for example, in a clinical study without the tester knowing the training settings (blind study). Thus, the training settings can be prepared in advance, and they do not have to be created during the test, as is the case with most currently available test equipment. In addition, neither the tester nor the hearing impaired burden with the test procedure.

The test or training can be documented with a result log. This includes (the logatoms who had to learn the most difficult), for example, the percentage of all ^¬ set-understood parts of speech (logatoms) and the Ziellogatome. In addition, the protocol may also include a common confusion matrix with a juxtaposition of presented and detected sounds. The results of the test can be an indicator of the extent to which the hearing aid has improved speech intelligibility. The result of the test over it ^¬ out can also be an indicator of the Be a training success. This may reduce the number of tests during a training session.

The individual training levels can be performed with and without additional noise. As a result, a direct comparison of the results can be achieved (speech intelligibility improvement in noise compared with improved speech intelligibility at rest). In addition, this comparison allows a speech intelligibility test of phonemes that are very sensitive to noise (target noise phenomens).

Claims

claims

A method for automatically training the speech comprehension of a person (10) wearing a hearing device (13)

a) acoustic presentation of a first language component (12) and

b) identifying (15) of acoustically presented speech component by the hearing apparatus (13) carrying Per ^¬ son (10),

marked by

c) automatically changing the acoustic performance of the

presented speech component and repeating (Sl to S4) of steps a) and b) with the altered performance if the identification is false until a predetermined maximum number of repetitions is reached, and d) acoustically presenting a second language component if the identifying (15) of the first language component is correct or the number of false identifications of the first language component is one greater than the maximum repetition number.

2. The method of claim 1, wherein the language component (12) is a logatom or a word.

3. The method of claim 1 or 2, wherein a number of language components (12) is specified and steps a) to d) are repeated until all voice components are presented at least once.

4. A method according to claim 2 or 3, wherein in step c) the altering consists of presenting (24), different emphasis or other background noise (23) to the respective previous performance with another voice.

5. The method according to any one of the preceding claims, wherein the language component (23) at the beginning of the method a Loga- tom and at its last repetition is a word in which the logatom is integrated.

A method according to any one of the preceding claims, wherein said identifying (15) is performed with a graphical user interface.

7. The method according to any one of the preceding claims, wherein in a false identification of the presented and the person (10) specified language component (12) are reproduced acoustically and / or visually.

8. The method according to any one of the preceding claims, wherein the person (10) always receives the voice component (12) at a constant volume from the hearing device (13).

9. The method according to any one of the preceding claims, wherein all process parameters by a trainer set beforehand and sent to the person to be trained by the trainer.

10. A device for automatically training the speech understanding of a person (10), which carries a hearing device (13) with

a) a playback device (11) for the acoustic presentation of a first language component (12) and

b) an interface device for inputting an identifier for identifying (15) the acoustically presented voice component (12) by the person (10) carrying the hearing device (13),

marked by

c) a control device that controls the display device (11) and the interface device so that

the acoustic performance of the speech component (12) is changed automatically and the steps a) and b) with the changed performance are repeated (S1 to S4) if the identification (15) is false, as often as until a predetermined maximum repetition number he ^¬ reaches, and

- A second language component is presented, if the identifying (15) of the first language component is correct or the number of false identifications of the first language component by one greater than the maxima ^¬ le repetition number.