Method for operating a hearing aid device and hearing aid device

Abstract

The invention relates to a method for operating a hearing aid device (2), the hearing aid device having an ear canal microphone (6) for collecting sounds in an ear canal (8) of a wearer and a own speech recognition unit (14) for implementing adaptive own speech recognition. In the method, the sounds in the ear canal (8) are collected by the ear canal microphone (6), and a signal (S) corresponding to the sounds is output by the ear canal microphone (6). G ), and only based on the signal (S) output by the ear canal microphone (6) G ) controls the training of adaptive own speech recognition. In addition, the invention also relates to such a hearing aid device (2).

Description

Technical Field

[0001] The present invention relates to a method for operating a hearing aid device having an ear canal microphone, wherein the sound in the ear canal of a wearer is collected by the ear canal microphone and a signal corresponding to the sound is output by the ear canal microphone. In addition, the present invention also relates to such a hearing aid device. Background Art

[0002] Hearing aids, in particular for hearing-impaired persons, are available in different designs to meet the individual needs of the wearer. For example, hearing aids are designed as behind-the-ear hearing devices (HdO, behind-the-ear “BTE”), hearing devices with external earpieces (receiver in channel “RIC”) or in-the-ear hearing devices (“IdO”).

[0003] The hearing aid device has at least one electroacoustic converter, in particular a microphone, which collects (air) sound from the wearer's surroundings. The corresponding signal output by the converter is processed, in particular amplified, by a signal processing unit in accordance with the wearer's needs. The processed signal is guided to the earpiece (speaker) and output from the earpiece as sound to the wearer's ear canal or eardrum.

[0004] The wearer (user) of the hearing aid perceives his own speech as different from the own speech received by the microphone of the hearing aid, which usually picks up air sound, due to structure-borne sound and / or (if the hearing aid partially or completely closes his ear canal) due to the occlusion effect.

[0005] For making the wearer normally perceive his own voice and external audio source, therefore need to carry out different processing to the wearer's own voice and other external audio sources.For this reason, first must identify the wearer's own voice.

[0006] For example, a method for recognizing the wearer's own voice is known from EP 3 451 705 A1. In the method described there for quickly recognizing the own voice of the wearer of a hearing aid, audio signals from at least one first microphone and a second microphone are used. The audio signal of the first microphone is filtered using a first filter, which attenuates the own voice of the wearer of the hearing aid. In addition, the audio signal of the second microphone is attenuated using a second filter, which attenuates external audio sources. The output signals of the two filters are compared with each other, and the voice of the wearer of the hearing aid is recognized based on the result. The first filter can be an adaptive filter. For example, the first filter is implemented by an adaptive algorithm, which adaptively determines the filter parameters of the first filter, so that these filter parameters can be determined when training with the wearer's own voice. In this case, the training is carried out within the scope of a conversation with the hearing device acoustician or by the wearer himself during his normal speech activity, wherein, for example, additional reference information about the existing speech activity is transmitted by a remote control or a smartphone.

[0007] Furthermore, DE 10 2016 203 987 A1 discloses a method for operating a hearing device, wherein the noise received by means of a microphone is analyzed with respect to its correspondence with the speech of a person wearing the hearing device. As a result of the analysis, a characteristic value is output which is compared with a threshold value. In this case, the threshold value is set depending on the environment, i.e., according to a certain environment. The determination of the threshold value for the corresponding environmental conditions takes place in a special calibration operation within the scope of a conversation session with an acoustician or in normal operation of the hearing device.

[0008] DE 10 2005 032 274 A1 discloses a hearing device and a method for operating the same. To detect the own speech of a person wearing the hearing device, a sound signal of a first microphone for receiving ambient sound and a sound signal of a second microphone for receiving ear canal sound are analyzed. For this purpose, the levels of the two sound signals are compared with one another and the presence of the own speech is inferred from the comparison.

[0009] EP 3 005 731 B1 describes a method for operating a hearing device comprising an ambient microphone, a signal processing unit, an earpiece and an ear canal microphone. Here, when the hearing device is switched on and worn in the ear canal of a user, the audio signal of the ambient microphone processed by the signal processing unit with a filter is filtered with a transfer function, which includes at least one transfer function from the output of the earpiece to the output of the ear canal microphone. The difference between the audio signal received by the ear canal microphone and the filtered signal is calculated, and the presence of the user's own voice is detected based on the difference.

[0010] From US 2019 / 0075406 A1 a hearing device is known, which has an ear canal microphone that can be worn on or in the ear canal of a user and a further behind-the-ear microphone arranged on or behind the ear of the user. In this case, the user's own speech is recognized when the sound pressure level at the ear canal microphone is at least 2.5 dB higher than the sound pressure level at the behind-the-ear microphone. The hearing device has a beamformer-filter unit from an adaptive algorithm to provide a spatially filtered signal. In this case, the beamformer-filter unit can include a predefined and / or adaptively updated own speech beamformer that is focused on the user's mouth. Summary of the invention

[0011] The technical problem to be solved by the present invention is to specify a method for operating a hearing aid, in which the wearer of the hearing aid is trained to recognize his own speech as effectively and / or as comfortably as possible. In addition, such a hearing aid should be specified.

[0012] With regard to the method, the technical problem is solved according to the invention by the features of the invention. With regard to the hearing aid, the technical problem is solved according to the invention by the features of the invention. Advantageous expansions and designs are the content of the invention. Here, by definition, the embodiments in connection with the method also apply to the hearing aid and vice versa.

[0013] The method is used to operate a hearing aid device. Such a hearing aid device is understood here to be a device for taking care of a hearing-impaired or hearing-impaired person, who expediently wears the device continuously or most of the time in order to compensate for the hearing defect. For example, the hearing aid device is designed as a behind-the-ear hearing device or an in-the-ear hearing device.

[0014] The hearing aid device has an ear canal microphone for collecting sounds, especially air sounds, in the ear canal of the wearer (user) of the hearing aid device. In particular, the ear canal microphone is arranged and designed to be placed in the ear canal of the user or worn there when the hearing aid device is used. If the sound in the ear canal is collected by the ear canal microphone, the ear canal microphone outputs a signal corresponding to the sound, that is, a signal representing the sound. This signal is also referred to as an ear canal signal.

[0015] The hearing aid device preferably further comprises a second microphone, the ambient microphone, for collecting sounds from the wearer's surroundings. When the sounds from the wearer's surroundings are collected by the ambient microphone, the ambient microphone outputs a further signal corresponding to the collected ambient sounds in a similar manner to the ear canal microphone, which is then also referred to as the ambient signal.

[0016] Furthermore, the hearing aid device has a self-speech recognition unit. The self-speech recognition unit is, for example, part of the control unit or, alternatively, is designed as a unit independent of the control unit. For example, the self-speech recognition unit is implemented as an integrated circuit. The ear canal signal and / or the ambient signal are guided from the ear canal microphone or from the ambient microphone to the self-speech recognition unit or the control device.

[0017] The own voice recognition unit is used to perform an analysis of the ambient signal and / or the ear canal signal, which is called own voice recognition or own voice detection (English: “own voice detection”, OVD), in order to determine whether the wearer's own voice is present.

[0018] The analysis of the corresponding signal is carried out in a suitable manner according to an algorithm or according to a filter or filters. In particular, the algorithm, the filter or one of the filters is designed so that when it is used, the signal corresponding to the wearer's own speech is attenuated as strongly as possible. Thus, during the analysis of the ambient signal and / or the ear canal signal, the presence of the own speech is inferred from the amount of attenuation. That is, in this case, the sound path from the wearer's mouth to the hearing aid device is mapped as accurately as possible and the transfer function corresponding to this sound path, i.e. corresponding to this sound path, is used to attenuate the signal.

[0019] Here, the own speech recognition is adaptive. In other words, the algorithm or one or more filters can be matched and changed, according to which the analysis of the corresponding signal is performed. In particular, one of their parameters or one of their coefficients is matchable and changeable. When training the own speech recognition, the algorithm or filter is set, in particular, their parameters and / or their coefficients are matched. The algorithm or filter is preferably set so that the part of the analyzed signal corresponding to a part of the wearer's own speech is attenuated more strongly than before training. Alternatively to this, the algorithm or filter is set so that the attenuation of the part of the analyzed signal corresponding to a part of the wearer's own speech, in particular the amount of attenuation, is adapted to a preset value, or is lower than a preset (and, for example, adjustable) threshold.

[0020] According to the method, such training of adaptive own speech recognition is controlled depending on the (ear canal) signal output by the ear canal microphone, in particular depending on its characteristics. In particular, the training is started, ended or interrupted depending on the (ear canal) signal. If the training is started, the filters or algorithms that have been matched by the previous training can also be further matched. If the ear canal signal or (particularly preferably only) the parameters derived from the ear canal signal have predefined characteristics, then the training is preferably started or the training that has already started is continued.

[0021] The invention is based on the finding that it is possible to determine particularly reliably whether the detected ear canal signal and, if applicable, the ambient signal are suitable for training (only) based on the ear canal signal. This allows for particularly reliable and efficient training.

[0022] Preferably, in addition to the ear canal signal, no further signals received by a further microphone, in particular ambient signals, are taken into account in order to determine how the training is to be controlled.

[0023] According to a suitable embodiment of the method, the ear canal signal is analyzed for the presence of own speech, and if own speech is recognized, training is started or (if training has already started) continued. If own speech is not recognized, training is not started, or a training that has already started is ended or interrupted (paused). In this case, it is relatively reliable to determine whether the wearer of the hearing aid device has his own speech, that is, whether the wearer himself speaks, based on the ear canal signal. Therefore, a particularly reliable indication of the presence of own speech can be determined based on the signal.

[0024] According to a suitable design, in order to analyze the ear canal signal for the presence of own speech, the level of the ear canal sound or the corresponding value of the ear canal signal is compared with a preset threshold value, wherein different threshold values ​​can be preset for different spectral ranges of the sound or signal, i.e. different frequency bands. If the level or the ear canal signal exceeds the threshold value, then own speech is inferred and training is started or continued.

[0025] Alternatively or additionally, in order to analyze the ear canal signal for the presence of own speech, the ear canal signal is examined for features that characterize the wearer's own speech using a spectral analysis, which features are produced, for example, by the sound path from the mouth through the head to the wearer's ear canal. For this purpose, the absorption is estimated or determined, for example, as a function of frequency.

[0026] As a further alternative or in addition, in order to analyze the auditory canal signal with regard to the presence of native speech, an algorithm or a filter of speech recognition is applied to the auditory canal signal and its attenuation is compared with a predefined threshold value.

[0027] According to an advantageous embodiment of the method, the training is started and carried out or continued automatically and thus online during normal operation. Normal operation is understood to be an operating mode of the hearing aid device, which is characterized in particular by the conventional use of the hearing aid device for compensating the hearing impairment of the wearer by processing the sounds received by at least one of the microphones and by outputting corresponding output sounds to the ear canal of the wearer. In particular, no previously defined speech sequences and / or noise sequences occur for the purpose of training the native speech recognition.

[0028] For the training, in particular for the matching of the algorithm or the filter or filters during the training, it is not necessary to resort to a protocol-defined process, which includes a preset speech sequence. Thus, the training process within the scope of a normal conversation (matching session) on the part of the acoustician or within the scope of a protocol-defined process by the wearer himself can be omitted. However, the training can also be carried out during a conversation session (matching session) on the part of the acoustician during a normal conversation, which therefore does not have a previously defined speech sequence and / or noise sequence.

[0029] Furthermore, it is not necessary to transmit additional reference information, in particular by the user through his current speech activity via an interface (eg a smartphone or a remote control) in order to start the training. In summary, the training of the native speech recognition takes place particularly conveniently for the wearer.

[0030] According to an advantageous embodiment of the method, it is determined based on the ear canal signal and / or based on the environmental signal whether the acoustic environment of the hearing aid device is suitable for training. For this purpose, in particular, the acoustic environment is analyzed with respect to noise and / or with respect to reverberation time. If the environment has been determined to be suitable, then the training is started or further carried out (continued) according to the result of the analysis, otherwise the training is ended or preferably interrupted until the environment is determined to be suitable again.

[0031] For example, a threshold value for noise values, in particular for noise levels, is preset, wherein training is started only when the threshold value is lowered. Alternatively, a signal-to-noise ratio, referred to as SNR, is determined in a suitable manner for the ear canal signal and / or the ambient signal and compared with a preset threshold value. Here, the threshold value for the signal-to-noise ratio is, for example, 15 db (decibel), 20 db or 25 db. When the signal-to-noise ratio determined from the corresponding signal exceeds the threshold value, training can be started or continued, and training is also started or continued appropriately.

[0032] In particular, a noise value, in particular a noise level in the acoustic environment of the hearing aid device, is determined for the ear canal signal by means of an interference noise estimator, a so-called "noise estimator", which is based, for example, on a "minimum statistics method". For this purpose, for example, the ear canal signal itself is taken into account. In order to determine the noise value more precisely, the ear canal signal is first amplified by a housing carrying the ear canal microphone and / or by a dome, based on a previously known attenuation of the ambient sound, and the noise value of the amplified signal is determined by means of the interference noise estimator.

[0033] The noise value of the ambient signal is also expediently determined by means of the interference noise estimator. Preferably, the training is started or continued only when the noise value of the ear canal signal and the noise value of the ambient signal are below a preset or respectively assigned threshold value.

[0034] Alternatively or additionally to the determination of the noise value, the acoustic environment of the hearing aid device is analyzed with respect to the reverberation time. If the determined reverberation time is less than a predefined threshold value, for example less than 500 ms, in particular less than 300 ms, then training can be started or further carried out and training can also be started or continued as appropriate.

[0035] In this way, inappropriate or even incorrect training, in particular inappropriate or incorrect adaptation of a filter or prefilters or algorithms of the native speech recognition, is advantageously avoided or at least the risk thereof is reduced, so that the training is more effective.

[0036] When the own voice is recognized based on the ear canal signal, for example based on its level or based on its spectrum analysis as described above, the acoustic environment of the hearing aid is preferably analyzed. Then, depending on the result of the analysis of the acoustic environment, training is started or further implemented or ended or interrupted until the environment is determined to be suitable again.

[0037] According to a suitable extension of the method, during or after the training, after the matching algorithm or the filter or filters, a convergence value (recognition value) is determined in a suitable manner. The convergence value represents a measure or quality of the recognition of own speech by means of own speech recognition. In addition, further training is permitted or training is continued or omitted depending on the convergence value.

[0038] For example, when using a filter or filters or algorithms for own speech recognition, the amount of attenuation of the ambient signal and / or ear canal signal is used as the convergence value. In this case, the ambient signal or ear canal signal essentially corresponds to the signal of the own speech, since, as mentioned above, training can only take place in a suitable acoustic environment, depending on the ear canal signal and therefore especially in the presence of own speech.

[0039] Suitably, the convergence value is compared with a preset threshold value, which corresponds in a suitable manner to an attenuation of 15 db or preferably 20 db or alternatively 25 db. If the convergence value is greater than the threshold value, i.e. if the quality of the recognition of the own speech is sufficiently high, further training is not allowed or the training is terminated. If the convergence value is less than the threshold value, further training is allowed or the training is continued in a corresponding manner. For example, further training is performed at a minimum time interval from the previous training. Alternatively, further training is performed as soon as the own speech is recognized again based on the ear canal signal.

[0040] According to an advantageous embodiment of the method, an incorrect positioning of the ear canal microphone in the ear canal and / or the ambient microphone in or on the ear is determined. If such an incorrect positioning has been identified, further training of the adaptive native speech recognition is started.

[0041] In this case, further training is preferably started based on the ear canal signal, in particular only when the wearer's own voice is recognized based on the signal in the above-mentioned variants and / or when the acoustic environment of the hearing aid device is suitable for further training.

[0042] For example, when placing or wearing glasses, hats or scarves, such mislocalization occurs. In this case, mislocalization is understood to be a deviation of the position of the ear canal microphone and / or the ambient microphone from the position for which the native speech recognition was trained.

[0043] For example, due to the mislocalization, the sound path from the wearer's mouth to the corresponding microphone is changed. As a result, there is a risk that one or more filters or algorithms of the own speech recognition do not correctly represent the sound path, and errors in the recognition of the own speech or incorrect determination of the noise value may occur. Advantageously, by further training after the recognition of the mislocalization, the algorithm or one or more filters are adapted again to the new position and / or orientation of the corresponding microphone. As a result, the risk of incorrect detection of the own speech or incorrect determination of the noise value is also at least reduced.

[0044] In a suitable embodiment of the method, a motion sensor is used to determine an incorrect positioning of the ambient microphone. The motion sensor is preferably arranged in or on a housing holding the ambient microphone. In a suitable manner, the orientation of the housing, in particular the position and / or the orientation of its axes, can be determined using the motion sensor. In the case of a rigid fastening of the motion sensor to the housing that firmly holds the ambient microphone, a corresponding change in the positioning or orientation of the ambient microphone can be achieved in an advantageously simple manner depending on a change in the positioning and / or orientation of the housing.

[0045] For example, an incorrect positioning of the ear canal microphone is determined by analyzing the ear canal signal. To this end, in particular, a spectral analysis of the signal is performed, the signal is checked for changes in features that characterize the native voice, and / or a change in the transfer function is checked.

[0046] According to the invention, a hearing aid has an ear canal microphone and a native speech recognition unit for implementing adaptive native speech recognition. The hearing aid is configured and designed to implement the method in one of the above-described variants. In particular, the native speech recognition unit is designed to start training for adaptive native speech recognition based on a signal output by the ear canal microphone. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The embodiments of the present invention will be described in detail below based on the accompanying drawings.

[0048] Figure 1A hearing aid device is schematically shown, which has an ear canal microphone, an ambient microphone and an own speech recognition unit; and

[0049] Figure 2 The method sequence for operating a hearing aid device is shown, in which the training of the intrinsic speech recognition is started as a function of a signal output by an ear canal microphone.

[0050] Parts and variables that correspond to one another are always provided with the same reference symbols in all the figures. DETAILED DESCRIPTION

[0051] Figure 1 The hearing aid device 2 shown has an ambient microphone 4, by which sounds from the surroundings of the wearer of the hearing aid device 2 can be picked up. In addition, the hearing aid device 2 has an ear canal microphone 6, which is used to pick up sounds in the ear canal 8 of the wearer. For this purpose, when the hearing aid device 2 is worn by the wearer as intended, the ear canal microphone 6 is inserted into the ear canal 8 of the wearer. In this case, the ear canal 8 is located at the ear canal 8 of the wearer. Figure 1 The lines are only schematically shown as mutually parallel lines.

[0052] according to Figure 1 The embodiment of the hearing aid 2 shown is designed as a so-called behind-the-ear hearing device. Here, the ambient microphone 4 is arranged in a housing 10 which is worn behind the ear, in particular between the pinna of the wearer and the wearer's head.

[0053] According to an alternative not shown in detail, the hearing aid device is designed as an in-ear hearing device, wherein ambient microphone 4 and ear canal microphone 6 are arranged in a common housing which is designed and configured for at least partial insertion into the ear canal.

[0054] The ambient microphone 4 and the ear canal microphone 6 are electroacoustic converters. They convert the sound collected in the wearer's surroundings (also referred to as ambient sound) or the sound collected in the ear canal 8 (also referred to as ear canal sound) into an ambient electrical signal S U Or ear canal electrical signal S G .

[0055] The microphones 4 and 6 are connected to a control unit 12 in terms of signal transmission, in which an own speech recognition unit 14 is integrated. Therefore, the own speech recognition unit 14 is a component of the control unit 12 .

[0056] Furthermore, the hearing aid device 2 has a receiver 16 (in this case a micro-speaker), which is likewise inserted into the auditory canal 8. The receiver 16 is likewise connected to the control unit 12 in terms of signal transmission, so that the receiver signal S H The control unit 12 can output the receiver signal S to the receiver.H The sound is converted into sound and output to the ear canal 8 of the wearer.

[0057] The hearing aid 2 further comprises a motion sensor 18 designed as an acceleration sensor, by means of which the position and / or orientation of the hearing aid can be determined. The motion sensor 18 is accommodated in the housing 10 and rigidly connected thereto, in which the ambient microphone 4 is also arranged.

[0058] The own speech recognition unit 14 is designed to carry out own speech recognition. In other words, the own speech recognition unit 14 detects the presence of the wearer's own speech in the ambient signal S U and / or ear canal signal S G Here, the environmental signal S U and / or ear canal signal S G The processing is performed by the control unit 12 or the own voice recognition unit 14, wherein the processing depends on whether the wearer's own voice has been recognized. As a result of the processing, an earpiece signal S H And output it to the earpiece 16.

[0059] Here, according to the signal S U and S G The own speech is recognized by means of own speech recognition with the aid of a certain number of filters F, i.e. one filter or more than one filter. The filter or filters F are adaptively configured here. The filter is thus changeable or adaptable, in particular during training. In summary, the own speech recognition is adaptive. The or at least one filter F is configured here so that when the filter is applied, the signal corresponding to the wearer's own speech or that part of the signal to be analyzed that corresponds to the own speech is attenuated as strongly as possible. Thus, the signal S analyzed with the aid of the filter F is U or S G The more it corresponds to the wearer's own voice, the more strongly the signal is attenuated.

[0060] According to an alternative not shown further, in a similar manner, the signal S U and / or S G The algorithm that performs the analysis is used to recognize own speech.

[0061] The own speech recognition unit 14 is designed to detect the speech of the ear according to the signal S output by the ear canal microphone 6. G Start training for adaptive self-speech recognition, according to Figure 2 This is shown in more detail.

[0062] exist Figure 2It can be seen from FIG. 1 that training is performed during normal operation N. In this case, the control unit 12 or the native speech recognition unit 14 in particular continuously processes the ear canal signal S G and / or environmental signal S U , and the handset signal S H The output is to earpiece 16 to compensate for the hearing impairment of the wearer.

[0063] The signal S output by the ear canal microphone 6 corresponds to or represents the sound in the ear canal 8 of the wearer. G It is directed to the control unit 12 , in particular the native speech recognition unit 14 (step I.).

[0064] In a second step II., the ear canal signal S is evaluated by the own speech recognition unit 14 based on the presence of own speech. G To this end, the ear canal sound is analyzed according to the ear canal signal S G The level P or ear canal signal S G The corresponding value of is compared with a preset threshold value. If it exceeds the threshold value, the wearer's own voice is considered to be known.

[0065] In addition, the ear canal signal S is analyzed for the presence of own speech. G Analyze the ear canal signal S with respect to at least one feature M that characterizes the wearer's own voice G Perform spectrum analysis.

[0066] Furthermore, in order to (redundantly) analyze whether the wearer's own voice is present, a filter F1 of a number of filters F of speech recognition is applied to the ear canal signal S G , and compare its attenuation with a preset threshold. If the attenuation is greater than the threshold, it is inferred that there is own speech.

[0067] According to an alternative not shown in further detail, the signal S G For analysis, only one or two of the methods shown above are considered, namely determining the signal S G level, its spectrum analysis or applying a filter to the signal S G .

[0068] The further method is preferably only carried out if the wearer's own voice was recognized in step II.

[0069] In a third step III. of the method, which follows the second step in time, the auditory canal signal S G To determine whether the acoustic environment of the hearing aid device 2 is suitable for training. To this end, the noise and the reverberation time t N Analyze the acoustic environment.

[0070] To this end, the ear canal signal S is determined G The signal-to-noise ratio SNR is calculated and compared with a preset threshold value. The noise is determined by means of an interference noise estimator.

[0071] According to an alternative not shown in detail, the ambient signal S U The signal-to-noise ratio SNR.

[0072] When the own voice is recognized according to step II., the ear canal signal S G When the signal-to-noise ratio (SNR) exceeds a preset threshold, the ambient signal S U When the signal-to-noise ratio (SNR) exceeds the threshold and the reverberation time (t) N When it is less than another preset threshold value, the training of the adaptive native speech recognition is started, or if the training has already started, the training is continued. During the training (step IV.), one or more filters of the native speech recognition are changed.

[0073] In a fifth step V., a convergence value K is determined, which is a measure of the recognition of the own speech by means of the own speech recognition. In this case, when using one or more filters F of the own speech recognition, the ambient signal S U and / or ear canal signal S G The amount of attenuation (weakening) of is considered as the convergence value K. If the convergence value K is less than another preset threshold value T K , then further training is allowed, which is Figure 2 In FIG. 1 , the arrow from step V. to step I. is shown. If the convergence value K is greater than the threshold value T K , then the training is ended and no further training is performed, that is, further training is omitted.

[0074] In a sixth step VI, based on the signal S of the motion sensor 18 B Determine the position of the ambient microphone 4. Here, the signal S B The motion sensor 18 outputs a signal to the control unit 12. The signal is evaluated by the control unit 12 with respect to its position and its orientation and therefore with respect to the position and orientation of the ambient microphone 4 and the housing 10. If an incorrect positioning is detected, i.e. the position and / or orientation of the ambient microphone deviates from the position at which the own speech recognition was previously trained, the method described above is repeated starting from step 1, wherein further training of the own speech recognition takes place. If the previously determined convergence value K is greater than the threshold value T K , then this also occurs. In this way, the risk of erroneous detection of one's own voice or erroneous determination due to incorrect localization is at least reduced.

[0075] In this case, it is automatically determined according to step VI. whether a faulty positioning is present during normal operation N. This is repeated regularly after a predetermined period of time, for example every 30 seconds.

[0076] In summary, according to the signal S output by the ear canal microphone 6 G The training of the adaptive native speech recognition is controlled. In this case, it is determined automatically, ie without user input, whether the training is to be started or continued.

[0077] The present invention is not limited to the above-mentioned embodiments. On the contrary, other variants of the present invention can also be derived therefrom by those skilled in the art without departing from the subject matter of the present invention. In addition, in particular, all individual features described in conjunction with the embodiments can also be combined with each other in other ways without departing from the subject matter of the present invention.

[0078] Reference numerals list

[0079] 2 Hearing Assistive Devices

[0080] 4 Ambient Microphones

[0081] 6 Ear canal microphone

[0082] 8 Ear canal

[0083] 10 Housing

[0084] 12 Control Unit

[0085] 14 Self-speech recognition unit

[0086] 16 Receiver

[0087] 18 Motion Sensor

[0088] I. Output ear canal signals to the self-speech recognition unit

[0089] II. Analysis of one's own voice

[0090] III. Evaluate the environment for hearing aids

[0091] IV. Training

[0092] V. Determining the convergence value

[0093] VI. Determining the Mispositioning of Ambient Microphones

[0094] F, F1 filter

[0095] K Convergence value

[0096] M Characteristics of own speech

[0097] N Normal operation of hearing aids

[0098] P Level

[0099] SNR Signal to Noise Ratio

[0100] S B Motion sensor signal

[0101] S G Ear canal signals

[0102] S H Handset signal

[0103] S U Environmental Signals

[0104] t N Reverberation time

Claims

1. A method for operating a hearing aid device (2), the hearing aid device having an ear canal microphone (6) for recording sounds in an ear canal (8) of a wearer and a native speech recognition unit (14) for carrying out adaptive native speech recognition, -in, The sound in the ear canal (8) is collected by the ear canal microphone (6), and the ear canal microphone (6) outputs an ear canal signal (S) corresponding to the sound. G ),and - wherein only the signal (S) output by the ear canal microphone (6) is used G ) starts training for adaptive self-speech recognition, - wherein said training is started automatically during normal operation, - in normal operation, the ear canal signal and / or the ambient signal are processed by the control unit and output to the receiver in order to compensate for the hearing impairment of the wearer, and In this case, during normal operation, no previously defined speech sequences and / or noise sequences are used for the purpose of training the own speech recognition.

2. The method according to claim 1, characterized in that - the effect of the presence of own speech on the ear canal signal (S G ) for analysis, and - if own speech is recognized, then start the training, or continue the training that has already started, or If own speech is not recognized, the training is not started, or a training that has already started is ended or interrupted.

3. The method according to claim 2, characterized in that In order to determine the presence of own speech, the ear canal signal (S G ) for analysis, - Determine the ear canal signal (S G ) level, - Checking the ear canal signal (S) according to spectral analysis G ) that characterizes the wearer's own voice, and / or - Applying a speech recognition filter to the ear canal signal (S G ).

4. The method according to any one of claims 1 to 3, characterized in that According to the signal (S G ) determines whether the environment is suitable for the training.

5. The method according to claim 4, characterized in that It is determined based on the signal (SG) whether the noise and / or the reverberation time in the environment is suitable for the training.

6. The method according to claim 4, characterized in that According to the signal (S) of the ear canal microphone (6) G ) to determine the noise value.

7. The method according to any one of claims 1 to 3, characterized in that - determining a convergence value (K) which represents a measure or quality of the recognition of own speech by means of own speech recognition, and - Depending on the convergence value (K), further training is allowed or omitted.

8. The method according to any one of claims 1 to 3, characterized in that - determining a mispositioning of the ear canal microphone (6) and / or the ambient microphone (4), and If such an incorrect localization has been detected, further training of the adaptive native speech recognition is started.

9. The method according to claim 8, characterized in that A motion sensor (18) is provided for determining a mispositioning of the ambient microphone (4).

10. A hearing aid device (2) having an ear canal microphone (6) and an own speech recognition unit (14), the own speech recognition unit being used to implement adaptive own speech recognition, wherein: The own speech recognition unit (14) is configured to detect a signal (S) outputted by the ear canal microphone (6) G ) starts training for adaptive self-speech recognition, - wherein said training is started automatically during normal operation, - in normal operation, the ear canal signal and / or the ambient signal are processed by the control unit and output to the receiver in order to compensate for the hearing impairment of the wearer, and In this case, during normal operation, no previously defined speech sequences and / or noise sequences are used for the purpose of training the own speech recognition.

Images