EP1523219B1 - Method for training and operating a hearing aid and corresponding hearing aid - Google Patents

Abstract

The training method has the hearing aid supplied with an acoustic input signal (2) with simulation of a number of hearing situations (3'), for matching the acoustic input signal to the hearing situation by the hearing aid user (1), the matching of the acoustic input signal to each hearing situation being learned automatically. Also included are Independent claims for the following: (A) a hearing aid provided with an automatic learning device; (B) an operating method for a hearing aid.

Description

The present invention relates to a method for training a hearing aid by providing an acoustic input signal, providing a plurality of hearing situation identifications and assigning the acoustic input signal to one of the hearing situation identifications by a hearing device wearer. Moreover, the present invention relates to a corresponding hearing aid that is nachtrainierbar, and a method for operating such a hearing aid after a night workout.

Classifiers are used in hearing aids to detect different situations. However, the default parameters may not necessarily be optimal for the appropriate situations with an individual hearing aid wearer. By night training, as it is usually used in speaker-related speech recognition systems, in certain situations, the recognition rate can be increased in terms of individual conditions. This is particularly important for the situation of the performance of one's own voice. Likewise, the classifier can be optimally adjusted to certain noise situations which are typical for the acoustic environment of the hearing device wearer.

In this context, from the document EP 0 681 411 A1 a programmable hearing aid known that adapts independently to changing environmental situations. The hearing device parameters are continuously adapted to the existing ambient noise, whereby "blurred" inputs can be used by the hearing aid wearer in addition to the measured input signals. The objective here is the parameters to optimize directly, whereby the hearing situation is not explicitly discussed.

In the publication EP 0 814 634 A1 Furthermore, a method is described with which the hearing device wearer optimally adjusts the hearing device by performing a night workout initiated by the hearing aid wearer. The hearing aid wearer receives a set of predefined parameter sets for the hearing situation that he communicates to the hearing aid for selection. From this limited set of parameter sets, each corresponding to a Hörgerätevoreinstellung, he selects the one that he feels to be optimal for himself. The corresponding hearing aid setting is learned by a set of rules, so that the same hearing aid setting takes place with a similar acoustic input signal. This means that the set of rules maps the acoustic input variables to the optimal hearing aid parameter set. In this night training, the hearing situation is only indirectly taken into account by the fact that only the parameter sets corresponding to this hearing situation are made available for selection. However, there is no direct adaptation of the hearing situation to the acoustic input data. A disadvantage is that the hearing aid wearer in such night training the sound of the hearing aid, which is determined by the parameter set used to assess. He has to judge, for example, whether he wants to get the sound brighter or darker. For certain complex algorithms and dynamic, adaptive parameters, such. For example, while controlling an adaptive directional microphone, differentiation between different sets of parameters is difficult, if not impossible, for the hearing aid wearer.

The object of the present invention is therefore to simplify the night training of a hearing device for the hearing aid wearer and to improve the operation of the hearing aid accordingly.

According to the invention, this object is achieved by a method for training a hearing aid by providing an acoustic input signal, providing a plurality of hearing situation identifiers and assigning the acoustic input signal to one of the hearing situation identifications by a hearing aid wearer and automatically learning the assignment of the acoustic input signal to the one of the hearing situation identifications.

Further, the invention provides a hearing aid with a recording device for receiving an acoustic input signal, a memory device for storing a plurality Hörsituationskennungen and an input device for assigning the acoustic input signal to one of the Hörsituationskennungen by a hearing aid wearer and a learning device for automatically learning the assignment of the acoustic input signal to the one the Hörsituationskennungen by the input device.

The invention is based on the knowledge that for the hearing aid wearer although the differentiation between different parameter sets is difficult, however, the naming of an acoustic situation that is currently present, for. As the situation "own voice" or "entertainment in the car", in most cases very reliable by the hearing aid wearer is possible. These situations go beyond the hearing situations conventionally used in hearing aids, such as "speech at rest" and "speech in noise". Ie. the more differentiated listening situations can relate to sub-aspects of these "classic" situations that are relevant for signal processing. The acoustic representations that this novel, Underlying more diverse situations can be easily retrained by specific naming. For example, the sound of one's own voice or the specific sound of one's own car can be learned by the hearing aid, for example through a neural network. The neural network thus takes contrary to the mentioned prior art according to EP 0 813 634 A1 a mapping of the acoustic input variables not to the resulting overall setting (parameter set) of the hearing aid, but to the internal situation representation (Hörsituationskennung) before. From this, the hearing aid parameter set to be used is then derived or the relevant parameters varied and / or supplemented based on audiological expert knowledge. In particular, the adaptive algorithms can reuse this information without the hearing aid wearer having to evaluate the result. This simple assignment of the input acoustic signal to given listening situations is far less difficult for the hearing aid wearer than the direct sound evaluation, such as the frequency response or compression ratios / knee points, according to the prior art because of the adaptivity of the algorithms and associated temporal dynamics.

In a specific embodiment according to the invention one of the listening situations of the performance of the own voice of the hearing aid wearer can correspond, so that after the automatic learning of the own voice can be detected. This is in many situations, for example, for the directional microphone setting of great importance.

The automatic learning of the at least one hearing device setting parameter for the assigned hearing situation on the basis of the automatic evaluation can take place during (online) or after (offline) the presentation of the acoustic input signal. In online night training, the acoustic input signal does not have to be completely stored, but the hearing aid requires more computing power for the night training perform. In offline night training eliminates this additional computing needs in the hearing aid, however, a memory device for the acoustic input signal is needed. The online evaluation avoids the time-consuming reading, processing and reprogramming of the data or the hearing aid.

The input device for assigning the acoustic input signal to a listening situation can also be used to start and stop the night training. As a result, the handling of the hearing aid or the implementation of the night training for the hearing aid wearer is simplified.

In addition, the input device may consist of a receiver integrated in the hearing device and an external remote control. The remote control may be configured for wired or wireless communication with the hearing aid. It is also conceivable that the remote control is used exclusively for the night training of the hearing aid. Alternatively, the remote control as a multi-function device, such as a mobile phone or portable computer with radio intersection speed, be configured.

The input device may also comprise a programmable arithmetic unit, in particular a PC, so that the operation takes place via a corresponding programming software.

Finally, in a specific embodiment, the input device can be verbally and in particular operated by means of one or more keywords. As a result, the operation of the hearing device for the hearing aid wearer is made even more comfortable.

Furthermore, the acoustic input signal may comprise a manually or automatically processed voice signal. This makes it possible to train the classifier very specifically.

When operating the hearing aid, d. H. After the night training, a currently valid parameter set can be influenced by automatically assigning the current hearing situation to a hearing situation identifier. In particular, a parameter of the parameter set can be varied and / or supplemented by the automatic allocation. Thereby, it is possible that the acoustic input signal is subjected to a complex signal processing based on expert knowledge, when the neural network is a learned hearing situation, for. B. own voice recognizes. In this case, the parameter set currently used in the hearing aid is changed accordingly or corresponding filtering is carried out.

FIG 1

ein Blockschaltbild zum Verfahren gemäß dem Stand der Technik;

FIG 2

ein Blockschaltbild des erfindungsgemäßen Verfahrens;

FIG 3

eine prinzipielle Darstellung eines Hörgeräts mit Fernbedienung zum Eingeben einer Hörsituation in einem ersten Schritt; und

FIG 4

die Situation des Hörgeräts gemäß FIG 3 während der Trainingsphase.

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

FIG. 1

a block diagram of the method according to the prior art;

FIG. 2

a block diagram of the method according to the invention;

FIG. 3

a schematic representation of a hearing aid with remote control for entering a hearing situation in a first step; and

FIG. 4

the situation of the hearing aid according to FIG. 3 during the training phase.

The embodiment described in more detail below represents a preferred embodiment of the present invention. For a better understanding of the invention, however, first the method for night training according to the prior art with reference to FIG. 1 again explained in more detail.

The hearing aid wearer or user 1 is located, as in FIG. 1 represented, in a special acoustic situation in the hearing aid, an acoustic input signal 2 available stands. Since the hearing aid for the hearing aid wearer 1 is subjectively not optimally adjusted, he undertakes a night training. For this he classifies the noise and tells the hearing aid the corresponding very general hearing situation or hearing situation identification, eg. As "speech in noise", with. Each of these listening situations 3 is assigned a plurality of parameter sets 4 in each case. Due to the selected listening situation 3, the hearing aid wearer 1 has, for example, seven parameter sets to choose from. He can now select the parameter set 4, with which the hearing aid is set so that it produces the subjectively best sound in this acoustic situation.

A neural network 5 learns the desired parameter set 4 for the applied acoustic input signal 2, so that it will again choose this parameter set 4 for a similar acoustic situation after the training phase.

The subjective assessment of the sounds caused by the different parameter sets for the hearing aid setting, however, is very difficult for the hearing aid wearer 1, since this requires a lot of detailed knowledge about the effects of the hearing aid parameters.

According to the present invention, therefore, the hearing aid should not train the use of special parameter sets, but only the recognition of the current situation. This is done according to the method of FIG. 2 , Here, too, the hearing aid wearer or user 1 receives the acoustic input signal 2. For the night training of the neural network 5 in the hearing aid, the hearing aid wearer 1 must assign the acoustic situation in which he is currently located to only one of a multiplicity of predefined, specific hearing situations 3 ' , The number of specific listening situations 3 'in the case of the present invention is usually greater than the number of general listening situations 3 according to FIG FIG. 1 because they should be differentiated from the beginning. Of the The reason for this is that the general listening situation "speech in noise" includes, for example, the specific hearing situation "own voice".

The neural network 5 does not learn the assignment of a parameter set to the acoustic input signal 2, but the assignment of a differentiated hearing situation or a hearing situation identifier 3 'to the acoustic input signal 2 (see arrows with solid lines in FIG FIG. 2 ). This means that the neural network, unlike the prior art, learns at a higher level. The example of the listening situation "own voice in your own car" this is explained in more detail. According to the prior art, this complex situation is assigned a fixed parameter set starting, for example, from the parameter set group "speech in noise". Since only a number of parameter sets for the selection for the hearing aid wearer for such situations "speech in noise" makes sense, certainly none of the available parameter sets for their own voice and additionally optimized for their own car.

According to the invention, however, the situation "own voice" and the other situation "in your own car" learned separately. These listening situations each have a specific influence on the complex signal processing. For example, in the situation "own voice" a specific amplification, possibly coupled with a special setting of the directivity of the hearing aid, and in the situation "in-house car" in turn causes a very specific noise reduction in the hearing aid.

It is particularly advantageous that one's own voice can be learned by the hearing aid. This takes place in that the acoustic input signal is subjected to a special processing with the own voice and specific parameters for the hearing aid are set specifically and assigned to the hearing situation "own voice". The same applies to learning, for example the hearing situation "own car", whereby a very specific noise reduction can be achieved. Thus, not only the input signal of a hearing situation is allocated here during learning, but also a very specific determination of parameters, such as filter or amplification parameters, takes place.

In use of the hearing aid after night training, the neural network 5 will assign an input acoustic signal 2 to one or more specific hearing situation identifiers 3 ', so that the currently valid parameter set 4' (including filter parameters) will be affected accordingly. A complex signal processing unit 6 z. B. with adaptive directional microphone, the signal processing on the basis of the influenced parameter set 4 'perform. If, according to the above example, the neural network now receives the input signal "own voice in own car", it assigns to it both the hearing situation identifier "own voice" and the hearing situation identifier "in own car", so that the current parameter set, for example with respect to the specific Amplification for your own voice and with regard to the specific filtering for the suppression of the noise of your own car is varied or supplemented.

In the following two concrete application examples of the present invention are shown:

Example 1: An adaptive directional microphone depends on the direction from which the maximum useful sound, z. As a voice signal, is incident. If the hearing aid wearer is talking to a person running next to him, the directional microphone should adjust to the conversation partner, ie to a maximum gain at an angle of approximately 90 °. However, as soon as the hearing aid wearer speaks for himself, the useful sound signal comes from his own mouth, ie from an angle of 0 °. The own language thus pulls the directional microphone characteristic away from the actual conversation partner and that usually with a certain delay. By contrast, if the hearing device is trained on its own voice, and thus the adaptive microphone control knows which acoustic properties belong to its own voice, signals which are classified as "own voice" can be disregarded for the tracking of the directional characteristic. In contrast, the adjustment of the hearing aid according to the prior art of EP 0 814 634 A1 according to FIG. 1 According to which the hearing aid wearer would have to assess several sets of parameters, promising little success due to the dynamics and adaptability of the processes. In particular, one's own voice could not be recognized.

Example 2: A noise suppression method can be specially trained for a complex time-varying noise. This noise is then optimally suppressed, although it may have similar spectral components or a modulation spectrum such as speech, which is to be further processed as a useful signal. Through individual training on this acoustic situation, z. For example, if the above-mentioned situation "in the car", the noise reduction method can be automatically set optimally, for example, by adjusting special weighting factors for individual spectral bands or the dynamic behavior is optimally adapted to the noise characteristic. Even in this case, the differences between the dynamic noise reduction settings are difficult to evaluate directly, while the situation is very reliable.

In certain acoustic situations, it may be advantageous if, in addition to the night training according to the invention, a night training according to the prior art takes place while the hearing aid wearer evaluates different parameter sets.

The night training, as it is for the hearing aid wearer, is now based on the 3 and 4 explained in more detail. The hearing aid wearer, for example, wants to train the situation "own voice" in his hearing aid 10. For this purpose, he connects via a line 11, a remote control 12 to the hearing aid 10. The remote control has a button 13 as a control.

In the classifier several listening situations are stored. The hearing aid wearer knows that the hearing situation "own voice" corresponds to situation 3, for example. Therefore, he presses the button 13 three times to tell the classifier that situation 3 should be retrained.

In a subsequent step, an acoustic signal, here the own voice, for recording for the hearing aid 10 according to FIG. 4 presented. The hearing aid wearer must now inform the hearing aid 10 about the beginning and the end of the training phase. This is done by holding the button 13 while speaking. This means that he must use only a single control element 13 for both steps of the training. If there are many listening situation identifiers, another design may be more user-friendly, e.g. B. with a display and a controller (slider, trackball, etc.), with the appropriate situation can be quickly selected.

The actual night training of the hearing aid 10 may occur during the performance of the acoustic signal 14. Alternatively, the acoustic signal 14 is recorded in the hearing aid and evaluated after recording or assigned to the selected auditory situation due to characteristic acoustic properties. In the case of online night training, a permanent or temporary storage of the acoustic signal 14 is not absolutely necessary.

Since the hearing aid 10 only the information about the current situation must be communicated, in contrast to the prior art according to EP 0 814 634 A1 an external operating unit is not absolutely necessary. It can, however, according to the 3 and 4 be used for example for reasons of comfort. However, it can also be a recording button attached to the hearing aid itself.

By night training, the recognition rate of the classifier for certain situations can be significantly increased compared to the default, so that the hearing aid sets more reliably to this situation. Due to the self-starting and ending of the night training phase by the hearing aid wearer situations can also be trained nachtrainiert reliable because the hearing aid wearer himself decides when the signal can be assigned to the situation.

Claims (24)

1. Method for retraining a hearing aid (10) by

   - provision of an acoustic input signal (2),

   - provision of two or more hearing situation identifications (3') and

   - association of the acoustic input signal (2) with one of the hearing situation identifications (3') by a hearing aid wearer (1),

   characterized by

   - automatic learning of the association between the acoustic input signal (2) and one of the hearing situation identifications (3').
2. Method according to Claim 1, in which one of the hearing situation identifications (3') identifies a hearing situation in which the hearing aid wearer's, own voice or the noise of his own automobile is presented, such that the wearer's own voice or the noise from his own automobile can be identified after the automatic learning process.
3. Method according to Claim 1 or 2, in which the learning process is carried out during the presentation of the acoustic input signal (2).
4. Method according to Claim 1 or 2, in which the learning process is carried out after the presentation of the acoustic input signal (2)
5. Method according to one of the preceding claims, in which the starting and stopping of the retraining and the association with the acoustic input signal (2) are carried out via a remote control (12).
6. Method according to Claim 5, in which the remote control (12) communicates with the hearing aid (10) without the use of wires.
7. Method according to one of Claims 1 to 4, with the hearing aid (10) being operated verbally for retraining.
8. Method according to Claim 7, in which the hearing aid (10) is operated by means of one or more keywords.
9. Method according to one of the preceding claims, with the acoustic input signal (2) comprising a speech signal which is preprocessed manually or automatically.
10. Hearing aid having:

    - a receiving device for receiving an acoustic input signal (2),

    - a storage device for storing two or more hearing situation identifications (3'), and

    - an input device (12) for the hearing aid wearer to associate the acoustic input signal (2) with one of the hearing situation identifications (3'),

    characterized by

    - a learning device (5) for automatically learning the association between the acoustic input signal (2) and one of the hearing situation identifications (3') by means of the input device (12).
11. Hearing aid according to Claim 10, in which one of the hearing situation identifications (3') identifies the hearing situation in which the hearing aid wearer's own voice or the noise of his own automobile is presented, such that the wearer's own voice or the noise from his own automobile can be identified after the automatic learning process.
12. Hearing aid according to Claim 10 or 11, in which the learning process of the learning device (5) can be carried out while the acoustic input signal (2) is being received in the receiving device.
13. Hearing aid according to one of Claims 10 to 12, in which the learning process can be carried out after the acoustic input signal (2) has been received in the receiving device, in the learning device (5) or in an external device with subsequent transmission to the hearing aid.
14. Hearing aid according to one of Claims 10 to 13, in which the input device (12) can be used for starting and stopping the recording or the retraining process.
15. Hearing aid according to one of Claims 10 to 14, with the input device (12) having an external remote control.
16. Hearing aid according to Claim 15, in which the remote control can be used for wire-free communication with the hearing aid.
17. Hearing aid according to Claim 15 or 16, in which the remote control is designed exclusively for retraining the hearing aid.
18. Hearing aid according to one of Claims 10 to 16, with the remote control being in the form of a mobile radio.
19. Hearing aid according to one of Claims 10 to 18, with the input device (12) comprising a programmable computation unit, in particular a PC.
20. Hearing aid according to one of Claims 10 to 19, in which the input device (12) can be operated verbally.
21. Hearing aid according to Claim 20, in which the input device (12) can be operated by means of one or more keywords.
22. Hearing aid according to one of Claims 10 to 21, in which the acoustic input signal (2) comprises a speech signal which is preprocessed manually or automatically.
23. Hearing aid according to one of Claims 10 to 22, having a signal processing device whose parameter set (4') can be influenced with the aid of the learning device (5) by association of the acoustic input signal (2) with a hearing situation identification (3') which has been learnt.
24. Hearing aid according to Claim 23, in which at least one parameter in the parameter set (4') can be varied and/or supplemented by the learning device (5) during the automatic association process.

Images