Hearing device and method for operating a hearing device for detecting the own voice on the basis of an individual threshold value

Abstract

The invention relates to a method for operating a hearing aid (2), wherein a sound is recorded by means of a microphone (4), the sound is analyzed with regard to its correspondence with the hearing aid wearer's own voice, and a feature value (M) is generated which indicates how closely the sound corresponds to the hearing aid wearer's own voice, wherein the wearer's own voice is a sound type (G1), wherein the feature value (M) is compared with a threshold value (S), wherein the sound is recognized as the wearer's own voice depending on whether the feature value (M) is above or below the threshold value (S), and wherein the hearing aid (2) is switched between several operating modes depending on whether the sound has been recognized as the wearer's own voice. The method is characterized in that the threshold value (S) is set according to the user.This results in improved voice recognition (10) which distinguishes the hearing aid wearer's own voice particularly reliably from another type of noise (G2). The invention further relates to a hearing aid (2) with corresponding voice recognition (10).

Description

[0001] The invention relates to methods for operating a hearing aid, wherein a sound is recorded by means of a microphone, wherein the sound is analyzed with regard to its correspondence with the hearing aid wearer's own voice and a feature value is generated which indicates how closely the sound corresponds to the hearing aid wearer's own voice, wherein the own voice is a noise type, wherein the feature value is compared with a threshold value, wherein the sound is recognized as the wearer's own voice depending on whether the feature value is above or below the threshold value, and wherein the hearing aid is switched between several operating modes depending on whether the sound was recognized as the wearer's own voice. The invention further relates to a hearing aid.

[0002] A corresponding method is described, for example, in the applicant's international application with the file number PCT / EP 2015 / 068796, which was published as WO 2016 / 078786 A1. By analyzing the sounds recorded by one or more microphones, it is possible to recognize the hearing aid wearer's own voice and, depending on this, switch the hearing aid between different operating modes. Such an analysis is also referred to as "own voice detection," or OVD for short. This is performed using an own voice recognition system, which is usually a component of the hearing aid. The microphone converts the sounds into electrical signals, which are then analyzed to assign the sound to a specific sound type, more specifically, to determine whether the original sound is the wearer's own voice or not, i.e.whether the hearing aid wearer is speaking or not.

[0003] US 2011 / 0261983 A1 describes a method for self-voice recognition in which a predetermined threshold for recognizing one's own voice is selected depending on ambient noise. For this purpose, different thresholds are first defined for different noise classes. During normal operation, i.e., when the hearing aid wearer is using the hearing aid, the threshold is selected depending on the currently present noise class.

[0004] In the application PCT / EP 2015 / 068796 cited at the beginning, the analysis is carried out using special filters, each of which has its own filter profile that is adapted to a particular noise, i.e. to a specific noise type or noise class. A given signal is then filtered using each of the filters. From the resulting filtered signal, it is then determined for each of the filters how strongly the original noise corresponds to the noise type to which a particular filter is adapted. The filter profiles are designed, for example, in such a way that the noise to be detected is attenuated as much as possible due to the filter profile. In the cited application, a distinction is made according to the location of the noise, i.e. noises that arise at different points in the room relative to the hearing aid are affected differently by a particular filter.This enables spatial differentiation and, in turn, differentiation of sound type based on its position relative to the hearing aid. Nearby sounds are recognized as spatially close and are then assumed to be one's own voice, while sounds further away are recognized as such and then assumed to be an unfamiliar voice. A closer match between the actual sound and the sound to which the filter is adjusted leads to greater attenuation and a higher level of agreement, i.e. a higher probability that the sound being examined corresponds to the sound type assigned to the filter. In this way, sounds can be correctly classified with a certain probability and assigned to one of, in particular, several different sound types.

[0005] Applying different filters to a recorded signal results in correspondingly different values ​​for the attenuation, i.e. generally the match, so that the type of noise can be determined based on these values. If the hearing aid wearer is speaking, the signal is attenuated more by this filter and the result is a higher value for the match than with another filter which, for example, is adjusted to an unfamiliar speaker in the frontal area of ​​the hearing aid wearer. By evaluating the two values, it can then be reliably determined that the hearing aid wearer is speaking, i.e. an own voice situation exists. The evaluation is carried out by forming a feature value, for example by calculating the difference or quotient of the two attenuation values, and then comparing the feature value with a predetermined, stored threshold or limit value.

[0006] US 2013 / 0148829 A1 describes a hearing aid with an analysis device which is designed to determine, depending on an audio signal, values ​​for a soft decision or for a probability as to whether a wearer is currently speaking.

[0007] US 2014 / 0088966 A1 describes a speech analyzer with a speech information collection unit and a distance calculation unit.

[0008] US Pat. No. 7,340,231 B2 describes a method for programming a communications device. A training session is conducted in which a user sets a signal processing parameter.

[0009] EP 2 381 702 A2 describes a method for a hearing aid user to use a hearing aid. A processor processes information from a database of the hearing aid user's speech, a speech feature, and a classification threshold to determine whether a noise-reduced signal is the hearing aid user's own voice.

[0010] Against this background, it is an object of the invention to provide a method for operating a hearing aid in which the wearer's own voice can be distinguished more reliably from other noises. Furthermore, a corresponding hearing aid with improved own voice recognition is to be provided.

[0011] The object is achieved according to the invention by a method having the features of claim 1 and by a hearing aid having the features of claim 10. Advantageous embodiments, further developments, and variants are the subject of the dependent claims. The statements made in connection with the method also apply mutatis mutandis to the hearing aid, and vice versa.

[0012] The method is used to operate a hearing aid. A hearing aid is generally understood to be a device for outputting sound, i.e. noises, via a loudspeaker, wherein the sound is obtained from noises picked up from the environment by at least one microphone. The noises are converted by the microphone into electrical signals and processed in the hearing aid by a control unit. The signals are then converted back into sounds via the loudspeaker and output. In particular, a hearing aid is understood to be a device for supplying a hearing-impaired or hard of hearing person who wears the hearing aid in particular continuously or most of the time in order to compensate for a hearing deficit. The hearing aid therefore has at least one microphone, a loudspeaker, also referred to as a receiver, and a control unit, the latter controlling the recording of sounds and their output.Usually, the control unit is designed at least to amplify noise.

[0013] In this method, a sound is recorded using a microphone. The sound, or more precisely the electrical signal generated from it, is analyzed for its similarity to the hearing aid wearer's own voice, and a feature value is generated that indicates how closely the sound matches the hearing aid wearer's own voice. The own voice is one type of sound among several different types of sounds.

[0014] The feature value is preferably generated using a classifier. A classifier analyzes a recorded sound with regard to a number of characteristic features of a specific sound type and provides the feature value as a measure of the match with the sound type. The feature value is then compared to a threshold value. Depending on whether the feature value lies above or below the threshold value, the sound is recognized as one's own voice, i.e., is clearly assigned to the "one's own voice" sound type. In this respect, the comparison with the threshold value is a decision-making process for determining at which value of the feature value the presence of one's own voice is assumed and when one's own voice is ultimately considered recognized.

[0015] The analysis of the noise, the generation of the feature value, the comparison with the threshold value and the decision as to whether or not the user's own voice is present are carried out by means of an own voice recognition system, which is a component of the hearing aid and can be implemented, for example, as an integrated circuit. The own voice recognition system can be part of the control unit of the hearing aid or designed as a separate unit. Depending on whether the noise has been recognized as the user's own voice, the hearing aid switches between several operating modes, for example an own voice mode and a non-own voice mode. The switching takes place automatically, i.e. by the hearing aid itself, in particular by the control unit or directly by the own voice recognition system.

[0016] According to the invention, the threshold value is set user-dependently and as an individual threshold value.

[0017] User-dependent determination of an individual threshold means that the threshold is set depending on the individual hearing aid wearer. In particular, no characteristic values ​​from other hearing aid wearers / users are used to determine the threshold.

[0018] The adjustment is made either during a fitting session with an audiologist, by the hearing aid wearer themselves, or during normal operation (i.e., online) and automatically by the hearing aid. By adapting the threshold used for comparison to the user, a potentially significantly different feature value is optimally accounted for when determining, and particularly classifying, one's own voice. It makes sense to also specifically adapt the generation of the feature value itself, as described above, to ensure particularly optimal recognition of one's own voice.

[0019] For the user-dependent, individual setting, the threshold is determined using a calibration procedure, in which the hearing aid wearer's own voice is recorded several times and several individual, i.e., user-specific, feature values ​​are generated. Finally, the calibration procedure sets the individual threshold based on the generated individual feature values. In this way, a particularly suitable and user-optimized threshold is set. Therefore, a large number of individual feature values ​​are generated, resulting in a distribution of the individual feature values, from which the threshold is then determined.

[0020] The threshold value is therefore set depending on the individual feature values ​​generated during the calibration procedure by setting the threshold value with respect to a characteristic value of the distribution, for example as a 2σ deviation from the mean value or generally in such a way that the generated feature values ​​are predominantly above or below the threshold value.

[0021] This design is based on the realization that the threshold can be highly user-dependent. Especially with the method described above and derived from PCT / EP 2015 / 068796, the attenuation values ​​generated by the filter used can vary significantly depending on the user. A fixed threshold would therefore result in one user's own voice being recognized and another user's being recognized as someone else's, even though in both cases the user's own voice is present.

[0022] Furthermore, this design is based on the consideration that both the user's own voice and external voices / ambient noises are recorded over time during the calibration process. Therefore, feature values ​​are obtained both in the presence of the user's own voice and in the presence of external voices / ambient noises. The overall distribution of feature values ​​thus obtained shows a range of possible feature values. From this distribution, the individual threshold is determined, for example, using statistical methods, in particular averaging.

[0023] This is based in particular on the realization that a feature value, which is determined and used to identify a sound and to assign it to a sound type, can vary greatly depending on the environment. In other words: in different environments for the hearing aid, a sometimes significantly different feature value may be generated when detecting a certain sound, for example, because it is recorded in an altered, distorted, or overlaid by other sounds. The term "environment" is to be understood in relation to the hearing aid and not to the hearing aid wearer. In particular, the hearing aid wearer's own voice logically differs from user to user, so that different hearing aid wearers also represent different environments for the hearing aid. But other sounds, i.e., external sounds with regard to the hearing aid wearer, e.g.foreign voices can lead to different characteristic values ​​in different environments.

[0024] "Noise" is generally understood to mean any type of sound signal in the audible frequency range. Different types of noise include one's own voice, another person's voice, tones, sounds, music, background noise, and hiss.

[0025] The method according to the invention is further based on the consideration that a decision regarding self-voice recognition based on a fixed threshold value is potentially highly error-prone. In order to reduce the error in determining the type of a noise, it is fundamentally possible to intentionally set the threshold value particularly high or particularly low. This can reduce the error rate in the erroneous recognition of noises other than one's own voice, or conversely, the error rate in the non-recognition of one's own voice despite it being present. Overall, however, this approach is inadequate because the correct recognition or non-recognition of one's own voice is limited to particularly clear cases, and the particularly environment-dependent value range of the feature values ​​is thereby largely excluded.

[0026] User-dependent setting of the threshold value is understood in particular to mean that the self-voice recognition does not use a generally predetermined threshold value for decision-making.

[0027] Rather, the appropriate threshold is selected, in particular, through a prior environmental analysis. For example, the self-voice recognition system itself or the control unit appropriately first determines the current environment and then selects and sets the corresponding threshold optimal for the environment from a group of thresholds.

[0028] A distinction must be made between the environment-dependent setting of the threshold during operation described above and a prior determination of the specific threshold to be used for this particular situation. This determination takes place either when adjusting the hearing aid, e.g. during a fitting session with an acoustician, or alternatively or additionally by the hearing aid wearer themselves. Automatic determination in a special calibration mode or during normal operation of the hearing aid is also conceivable in principle. In general, the determination creates an assignment of thresholds to environments, so that a group of thresholds is available for selection, from which the most suitable is then set. This assignment is expediently stored in a memory of the hearing aid, in particular the control unit, for example as a table, as a functional assignment or as a user profile.In this respect, not just one predetermined threshold is stored, but several predetermined thresholds for different environments. From these multiple predetermined thresholds, a suitable one is then selected and adjusted depending on the environment, which significantly reduces the risk of errors when selecting the hearing aid's operating mode.

[0029] The user-dependent setting of the individual threshold value must also be distinguished from setting the determination of a feature value, for example setting the filter mentioned above or a classifier which is used to analyze noises and generate a feature value. The threshold value therefore does not serve to determine the feature value, but rather to evaluate the already determined feature value. Such a configuration of the components which generate the feature values ​​takes place in particular independently of the user-dependent or environment-dependent selection and setting of the threshold value for the evaluation of the feature value. Nevertheless, these components are also expediently set in a user-dependent manner. This is useful, for example, with regard to self-voice recognition, i.e. the recognition of the voice of the hearing aid wearer, i.e. the generation of the feature value, for exampleby a filter, is appropriately adapted to the voice of the hearing aid wearer in order to ensure optimal feature value generation and thus optimal distinguishability from other types of noise.

[0030] In a suitable further development, the threshold is calibrated by determining a maximum and a minimum feature value over a limited period of time and setting the threshold between the minimum and the maximum feature value. This is based in particular on the assumption that at the maximum feature value the noise is of the "own voice" noise type and at the minimum feature value the noise is of the "other voice" noise type. Depending on the calculation of the feature value, however, this can also be the other way around, i.e. it is then assumed that the own voice produces a minimum feature value and the other voice a maximum feature value. The limited period of time is usually a few seconds to a few tens of seconds, for example, around 20 s. The maximum and minimum feature values ​​are thus short-term extremes within the period.By continuously determining short-term extremes, typical feature values ​​are determined for one's own voice and for another noise type, in particular a foreign voice, over a significantly longer period than the limited period. This advantageously yields statistical distributions at least similar to those obtained with the calibration procedure described above, in which at least the presence of one's own voice must be known. In this case, however, based on the minimum and maximum feature values ​​within the limited period, it is essentially a guess as to when the recorded noise is one's own voice and when it is another noise type.

[0031] In an advantageous embodiment, the threshold is calibrated during normal operation by repeatedly determining the individual feature values ​​and adjusting the threshold accordingly. This continuously adjusts the threshold so that the threshold values ​​stored during the assignment approach optimal thresholds over time.

[0032] Calibration does not correspond to the environment-dependent adjustment of the threshold, which is set in a specific situation. Rather, calibration involves adjusting the threshold stored for a specific value range, which is then adjusted. In this sense, the recurring recalibration of the threshold of a value range represents a continuous online optimization of self-voice recognition. This optimization occurs either continuously, only at specific times, or only over a single specific period.

[0033] In this case, in addition to the sound's match with the listener's own voice, the sound is also analyzed for a match with at least one other sound type. For example, a match value is generated which indicates how strongly the sound matches a certain sound type, with the match values ​​then being combined to form the feature value. One of the at least two sound types is the listener's own voice. This feature value then makes it possible to distinguish between the listener's own voice and the other sound type based on the feature value. This differentiation is significantly improved by the threshold value set depending on the environment. The feature value is, for example, the difference or the quotient of the two match values.

[0034] In a preferred variant, the distinction between one's own voice and another type of noise corresponds to the distinction between spatially separated noises. One's own voice is usually the type of noise that is spatially closest to the hearing aid, so spatial differentiation, i.e., differentiation based on the location of the noise, also makes it easy to distinguish between one's own voice and another type of noise.

[0035] In a preferred embodiment, the other noise type is a foreign voice, which is positioned particularly frontally with respect to the hearing aid wearer. In this case, a foreign voice is not understood to mean the voice of a specific other person, but rather a voice that is not the hearing aid wearer's own voice. Self-voice recognition then distinguishes between one's own voice and a foreign voice.

[0036] In a particularly preferred embodiment, the feature value is generated as in the international application PCT / EP 2015 / 068796 cited above using a pair of filters, one of which is configured to maximally attenuate the wearer's own voice and the other filter to maximally attenuate a foreign voice, in particular a foreign voice coming from a person directly in front of the hearing aid wearer. When analyzing a noise, the two filters each provide a match value, and the feature value is then calculated from the two match values, e.g. by subtracting the match value for the foreign voice from that of the wearer's own voice. The feature value is then lower for a foreign voice than for the wearer's own voice. If the threshold value is undershot, the noise is recognized as a foreign voice; if the threshold value is exceeded, however, the noise is recognized as the wearer's own voice.

[0037] The generation of feature values ​​for other noise types is often user-dependent. Therefore, during the calibration process, another noise type, particularly a foreign voice, is recorded before or after recording one's own voice. Here, too, several feature values ​​are generated, analogous to what was previously said, depending on which feature values ​​the threshold is set. This significantly improves calibration, particularly with regard to the accuracy of distinguishing one's own voice from the other noise type. The threshold is then set, for example, to the mean of the two mean values ​​of the two generated statistical distributions for the two noise types.

[0038] The person wearing the hearing aid is not the only environmental condition for which it makes sense to adjust the threshold. Of particular importance when analyzing most types of noise is their superposition with noise, often background noise or interfering noise. In particular, it has been recognized that generating a feature value, i.e. in particular classifying the noise, becomes more difficult and error-prone as the volume of the noise increases. The same applies analogously to distinguishing between two types of noise. Therefore, in a particularly preferred embodiment and alternatively or in addition to the user-dependent setting of the threshold, the threshold is set depending on the environment by determining a noise value and setting the threshold as a function of the noise value. This further optimizes self-voice recognition.

[0039] The noise value characterizes the noise and, in particular, quantifies it. The noise value is preferably a level, a volume, an intensity, or an amplitude of the noise. Alternatively, the signal-to-noise ratio is also suitable as a noise value. A typification of the noise is also suitable, i.e., the assignment of the currently present noise to a specific noise type and a setting of the threshold value depending on the detected noise type, with the noise type then being the noise value.

[0040] In addition to or as an alternative to the noise-dependent setting, any other environmental dependency is also suitable to be first determined and, in particular, quantified in order to then adjust the threshold value depending on it.

[0041] In a suitable embodiment, several value ranges are defined for the noise value, each of which is assigned a threshold value. The value range in which the noise value lies is then determined, and then the threshold value assigned to the determined value range is selected and set. In this way, a sufficiently suitable threshold value is easily assigned to each noise value, resulting in an overall assignment, e.g. in the form of a table, from which the most suitable threshold value for a particular situation is selected and then set. This is based on the idea that the noise value lies within a certain value spectrum, which is then advantageously divided into several, in particular connected, intervals in order to implement a noise-dependent setting of the threshold value.

[0042] For example, the noise level is the level of noise in the hearing aid's environment. The level is usually specified in dB. The value spectrum then ranges, for example, from -90 to -40 dB and is divided into approximately 10 to 20 value ranges of, for example, 5 dB each. Each value range is then assigned its own threshold. During operation of the hearing aid, the noise level is measured, and the threshold value assigned to the value range in which the measured level lies is then set. The level is measured, for example, using a noise estimator, e.g., based on a minimum statistics approach.

[0043] The assignment of threshold values ​​to value ranges takes place, for example, during a fitting session with an acoustician or by the hearing aid wearer themselves, e.g. as part of a calibration procedure. It is particularly important that defined noise values ​​are available or can at least be reliably measured. The assignment can be made using a pure calibration measurement and then presented as a table and saved on the hearing aid, or the assignment is made using a functional assignment, which is, for example, an approximation of the result of the calibration measurement. In the latter variant, for example, an upper and a lower limit is assumed for the threshold value, in particular an upper limit for low levels, e.g. below -75 dB, and a lower limit for high levels, e.g. above -60 dB, and a linear extrapolation is carried out in between.In this case, it is advantageous to only determine a suitable upper and lower limit, as well as the value ranges over which extrapolation is then carried out.

[0044] In a practical embodiment, the threshold is recurrently recalibrated during normal operation of the hearing aid, in particular as described above with regard to the user-dependent determination of the optimal threshold. The user-dependent threshold is thereby continuously calibrated and, over time, increasingly better adapted to the current hearing aid wearer. This corresponds, in particular, to a training mode for the hearing aid, which is expediently terminated after a certain training period. The user-dependent threshold is then, in particular, fixed.

[0045] The hearing aid according to the invention has an own voice recognition system, which is designed to carry out the method in one of the above-mentioned embodiments. Depending on the result of the own voice recognition, the hearing aid is then switched to a suitable operating mode for the respective situation. In one variant, the switching is also performed by the own voice recognition system.

[0046] Below, an example of implementation is explained in more detail using a drawing. It shows: Fig. 1 a hearing aid with an own voice recognition, Fig. 2 a graphic representation of the results of a measurement for the recognition of the own voice of a hearing aid wearer, Fig. 3 a graphic representation of the results of another measurement for the recognition of the own voice of a hearing aid wearer.

[0047] In Fig. 1 A hearing aid 2 is shown schematically. This is designed here as a so-called BTE device and is worn by a user behind the ear. In one variant, the hearing aid 2 is an ITE device and is worn in the ear. Other hearing aid types are also suitable in principle. The hearing aid 2 has a microphone 4 for picking up sounds from the environment of the hearing aid 2. A picked up sound is processed as a signal in a control unit 6 of the hearing aid 2 and prepared for output via a loudspeaker 8. The signal, i.e. the sound, is usually amplified.

[0048] The hearing aid further comprises an own-voice recognition system 10, which in the illustrated embodiment is part of the control unit 6. The control unit 6, the own-voice recognition system 10, the microphones 4, and the loudspeaker 8 are suitably connected to one another. Furthermore, the hearing aid 2 can be operated in various operating modes, between which the user can switch using the control unit 6 or the own-voice recognition system 10. The own-voice recognition system 10 analyzes the recorded sounds and assigns them to specific sound types G1, G2, for example, the sound type G1 "own voice" or the sound type G2 "foreign voice." Depending on the detected sound type G1, G2, the system then switches to a suitable operating mode. For recognition, the own-voice recognition system 10 generates a feature value M and compares it with a threshold value S to determine which sound type G1, G2 the analyzed sound belongs to.This is related to the . Fig. 2 and 3 described in more detail below.

[0049] The Fig. 2 and 3each show the results of a measurement in which a sound was recorded and analyzed several times in succession. Two different sound types G1, G2 were used: the hearing aid wearer's own voice on the one hand, and an external voice on the other. The own-voice recognition 10 of the hearing aid 2 first analyzes the recorded sound with the aim of assigning it a feature value M, which provides information about whether the sound belongs to one or the other sound type G1, G2. In the present case, this was achieved using a pair of filters with two filters that have different filter profiles. The filters are designed such that one filter attenuates the wearer's own voice as much as possible, and the other filter attenuates the external voice. A feature value M is generated by comparing the two different attenuations for the same sound.

[0050] The multitude of characteristic values ​​M, which were recorded during the measurements, are shown in the Fig. 2 and 3 and plotted against a noise value R, in this case the level of ambient noise. The noise value is given here in decibels (dB). The noise value R is measured, for example, using a background noise estimator. The feature values ​​M are also assigned to one of two groups, depending on which noise type G1, G2 was actually presented to the hearing aid. The feature values ​​M, which were generated during the analysis of one's own voice as noise type G1, are shown in light gray, and the feature values ​​M, which were generated during the analysis of the other person's voice as noise type G2, are shown in black. The measurements of the Fig. 2 and 3 differ in that these results show for different hearing aid wearers, ie at least one's own voice is different.

[0051] It is clearly visible in the Fig. 2 and 3 that when a foreign voice is present, a smaller feature value M is predominantly generated than when one's own voice is present. This makes it possible to define a threshold value S with which a specifically generated feature value M is compared in order to decide which noise type G1, G2 is present. In the exemplary embodiment, a noise is recognized by the own voice recognition system 10 as one's own voice if the feature value M is greater than the threshold value S, and as a foreign voice if the feature value M is less than the threshold value S.

[0052] Conventionally, only a fixed threshold S is used to be compared with the feature value M in any situation and environment. As can be seen from the Fig. 2 and 3becomes clear, however, this may be insufficient. Rather, it is evident that the use of different threshold values ​​S is useful in different environments. A first environmental dependency is that the generation of the feature value M is strongly dependent on the noise value R. For low noise values ​​R, comparatively large feature values ​​M are generated for one's own voice, but with a larger noise value R, the difference to the feature values ​​M of the foreign voice is significantly smaller. Therefore, a lower threshold value S is advantageously selected for larger noise values ​​R.

[0053] In Fig. 2 the optimal threshold values ​​S for individual value ranges W of the noise value R are entered, namely as gray horizontal bars. This effectively assigns a threshold value S to a specific value range W, resulting in an overall assignment Z1 in the manner of a table. The hearing aid 2 then determines a feature value M for a sound that has just been recorded and also the environment, in this case the noise value R, i.e. effectively the level or loudness of the noise that is superimposed on the noise. Before the comparison with the feature value M, the threshold value S is then adjusted depending on the environment, namely to the threshold value S that is assigned to the value range W in which the determined noise value R lies. In this way, the feature value M is compared with a threshold value S that is adapted to the given situation and an optimal result is achieved when distinguishing between one's own voice and that of another.

[0054] Instead of the table-like assignment Z1 of the optimal threshold values ​​S to the value ranges W, a simplified assignment Z2 is used. This is also shown in Fig. 2 represented as a dark gray, stepped line. For simplicity, it is assumed that below a low noise level Rmin, a maximum threshold Smax is sufficient, and above a high noise level Rmax, a minimum threshold Smin. In between, an extrapolation of the thresholds S takes place, here according to a linear relationship with regard to the selected representation. Overall, the simplified assignment Z2 effectively smooths the assignment Z1 with the optimal thresholds S. The assignment Z2 is saved in one variant as a simple table; alternatively, a function is saved for calculation.

[0055] When comparing the Fig. 2 and 3a further environmental dependence of the feature values ​​M becomes clear, namely the person of the hearing aid wearer. Fig. 3 are on the one hand as well as in Fig. 2 an assignment Z1 of optimal threshold values ​​S to certain value ranges W is shown as gray horizontal bars. In addition, the same simplified assignment Z2 from Fig. 2 in the Fig. 3 again as a dark grey, stepped line. When comparing the simplified assignment Z2, which is for the hearing aid wearer from Fig. 2 was determined, with the optimal threshold values ​​S for the other hearing aid wearer of the Fig. 3 According to the assignment Z1 it is immediately clear that the Fig. 2 determined assignment Z2 in Fig. 3 is not optimal. Therefore, the threshold value S is advantageously set user-dependently, i.e., depending on the individual hearing aid wearer.

[0056] Overall, the threshold value S is therefore preferably set in two ways depending on the environment: firstly, user-dependent, and secondly, depending on the noise level R measured at a given time. Which threshold value S is then specifically set—i.e., one or more of the assignments Z1, Z2—i.e., which threshold values ​​S are available for selection, is expediently determined in a calibration procedure. This is performed either during a fitting session with an acoustician, by the hearing aid wearer themselves, automatically by the hearing aid as part of online optimization, or a combination of these.

[0057] To determine an optimal threshold value S for a given hearing aid wearer and for a specific noise level R, the methods described above in connection with the Fig. 2 and 3described measurements. In this case, sounds of a known sound type G1, G2 are analyzed and the determined feature values ​​M are used as typical feature values ​​M to determine a suitable threshold value S. When using two different sound types G1, G2, for example, two different statistical distributions of feature values ​​M are determined and then a threshold value S is selected between them. However, it is also conceivable to use only one sound type G1, G2. In one variant, calibration is carried out by using previously known sound types G1, G2, so that the correct assignment is trained.In another variant, calibration is performed during normal operation of the hearing aid 2 by generating feature values ​​M in limited time periods of a few seconds to a few tens of seconds and under the assumption that the extremes of the feature values ​​M determined in a respective time period can be assigned with sufficient certainty to a specific noise type G1, G2. For example, it is assumed that the generation of a maximum feature value M was caused by the user's own voice and the generation of a minimum feature value M by another person's voice. These extremes are then used to determine an optimal threshold value S, which can be and expediently is further adjusted during subsequent operation of the hearing aid 2 through ongoing calibration. List of reference symbols

[0058] 2Hearing aid 4Microphone 6Control unit 8Speaker 10Own voice detection G1, G2Noise type MFeature value RRoise level RminLow noise level RmaxHigh noise level SStreshold SminMinimal threshold SmaxMaximum threshold WValue range Z1, Z2Assignment

Claims

1. A method for operating a hearing aid (2), - wherein the sound is recorded by means of a microphone (4), - wherein the sound is analyzed with respect to its correspondence with the own voice of the hearing aid wearer and a feature value (M) is generated which indicates how strongly the sound corresponds with the own voice of the hearing aid wearer, - wherein the own voice is a sound type (G1), - wherein the feature value (M) is compared with a threshold value (S), - wherein the sound is identified as the own voice depending on whether the feature value (M) is above or below the threshold value (S), and - wherein the hearing aid (2) is switched between multiple operating modes depending on whether the sound was identified as the own voice, - wherein the threshold value (S) is determined depending on the user and is set as an individual threshold value (S), in that the threshold value (S) is determined by means of a calibration method, in which the own voice of the hearing aid wearer is recorded and multiple individual feature values (M) are generated, and in which finally the individual threshold value (S) is set depending on the generated individual feature values (M), characterized in that in addition to the correspondence with the own voice, the sound is also analyzed with respect to a correspondence with at least one other sound type (G2) and the other sound type (G2) is recorded in the calibration method before or after the recording of the own voice, and multiple feature values (M) are also generated in this case, depending on which the threshold value (S) is set.

2. The method as claimed in claim 1, wherein in each case a correspondence value is generated, which indicates how strongly the sound corresponds with a specific sound type (G1, G2), wherein the correspondence values are then combined to form the feature value (M).

3. The method as claimed in claim 1 or 2, characterized in that the threshold value (S) is calibrated in that a maximum and a minimum feature value (M) are determined over a limited period of time and the threshold value (S) is set between the minimum and the maximum feature value (M).

4. The method as claimed in any one of the preceding claims, characterized in that the threshold value (S) is repeatedly recalibrated in a normal operation during the use of the hearing aid (2) by the hearing aid wearer.

5. The method as claimed in any one of claims 1 to 3, characterized in that the other sound type (G2) is a foreign voice, which is in particular arranged frontally in relation to the hearing aid wearer.

6. The method as claimed in any one of the preceding claims, characterized in that the generation of the feature value is carried out by means of a filter pair, wherein one of the filters is configured for a maximum attenuation of the own voice and the other filter is configured for a maximum attenuation of a foreign voice.

7. The method as claimed in any one of the preceding claims, characterized in that the threshold value (S) is set depending on the environment in that a noise value (R) is determined and the threshold value (S) is set depending on the noise value (R).

8. The method as claimed in the preceding claim, characterized in that multiple value ranges (W) are defined for the noise value (R), which are each assigned a threshold value (S), the value range (W) is determined, in which the noise value (R) lies, and the threshold value (S) is selected and set which is assigned to the determined value range (W).

9. The method as claimed in one of the two preceding claims, characterized in that the threshold value (S) is calibrated in normal operation in that the noise value (R) is repeatedly determined and the threshold value (S) is calibrated depending thereon.

10. A hearing aid (2) having a microphone (4) for recording a sound and having an own voice identification function (10), which is designed such that - the sound is analyzed with respect to its correspondence with the own voice of the hearing aid wearer and a feature value (M) is generated which indicates how strongly the sound corresponds with the own voice of the hearing aid wearer, - wherein the own voice is a sound type (G1), - the feature value (M) is compared with a threshold value (S), - wherein the sound is identified as the own voice depending on whether the feature value (M) is above or below the threshold value (S), - a switch is made between multiple operating modes depending on whether the sound was identified as the own voice, - the threshold value (S) is determined depending on the user and is set as the individual threshold value (S) in that the threshold value (S) is determined by means of a calibration method, in which the own voice of the hearing aid wearer is recorded and multiple individual feature values (M) are generated, and in which finally the individual threshold value (S) is set depending on the generated individual feature values (M), characterized in that the sound, in addition to the correspondence with the own voice, is also analyzed with respect to a correspondence with at least one other sound type (G2) and in the calibration method, the other sound type (G2) is recorded before or after the recording of the own voice, and multiple feature values (M) are also generated in this case, depending on which the threshold value (S) is set.

Images