EP2620940A1 - Method and hearing device for estimating a component of one's own voice - Google Patents

Abstract

The method involves segmenting a time-frequency plane into a number of regions, and determining a region phase difference and a region level difference respectively for each region from signals with respect to each other. The regions of the time-frequency plane are grouped such that the region phase difference corresponds to an estimated phase difference and the region level difference corresponds to an estimated level difference. The signal components of the group are acted as an estimation of a voice component of a wearer. An independent claim is also included for a hearing device.

Description

The present invention relates to a method of estimating a constituent of a listener's own voice. Moreover, the present invention relates to a hearing device in which a corresponding method is implemented. Furthermore, the present invention relates to a hearing device having a filter, which was created according to the above method. A hearing device is here understood to mean any device which can be worn in or on the ear and produces a sound stimulus, in particular a hearing device, a headset, headphones and the like.

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

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

In numerous hearing aid applications, it is necessary or desirable to be able to extract the speech or voice of the wearer of the hearing device or the hearing device from the sound environment. An exemplary application would be the active reduction of occlusion effects. In addition, a beamformer can be controlled by his own voice. Furthermore, it is conceivable that the space impulse response is estimated on the basis of one's own language.

The own language or speech components of the wearer of the hearing device can be estimated or extracted with different methods. A well-known method for doing so is known as CASA (Computational Auditory Scene Analysis). This CASA principle is therefore based on the fact that the current hearing situation is analyzed mathematically. The CASA principle is based on the ASA principle, its main achievements in the work of Bregman, AS (1994): "Auditory Scene Analysis: The Perceptual Organization of Sound", Bradford Books summarized , The state of developments regarding CASA is reproduced in the article Wang, D., Brown, GJ (2006): Computational Auditory Scene Analysis: Principals, Algorithms and Applications, John Wiley & Sons Publishing, ISBN 978-0-471-74109-1 ,

Monaural CASA algorithms work on a single signal channel and try to separate the sources. At least, language should be separated. In most cases, they are based on very strict conditions with regard to the sound sources. One of these conditions relates, for example, to the fundamental frequency estimate. In addition, the monaural CASA algorithms are in principle unable to use the spatial information of a signal.

Multi-channel algorithms attempt to separate the signals based on the spatial locations of the sources. In this approach, the configuration of the microphones is essential. For example, in binaural configuration, i. H. if the microphones are on both sides of the head but no reliable source separation is done with these algorithms.

The object of the present invention is to be able to recognize the own voice of a wearer of a hearing device more reliably.

• Platzieren eines ersten Mikrofons der Hörvorrichtung am Ausgang des Gehörgangs eines Ohrs des Trägers oder außerhalb des Gehörgangs,
• Platzieren eines zweiten Mikrofons der Hörvorrichtung in dem Gehörgang, so dass sich das zweite Mikrofon näher an dem Trommelfell des Ohrs befindet als das erste Mikrofon,
• Schätzen einer Phasendifferenz und einer Pegeldifferenz virtueller Mikrofonsignale der beiden Mikrofone zueinander anhand eines vorgegebenen Modells,
• Gewinnen je eines zeitlichen Mikrofonsignals durch jedes der beiden Mikrofone,
• Transformieren der beiden zeitlichen Mikrofonsignale zu je einem t-f-Signal in die Zeit-Frequenz-Ebene,
• Segmentieren der Zeit-Frequenz-Ebene in mehrere Regionen,
• Ermitteln jeweils einer Regionphasendifferenz und einer Regionpegeldifferenz für jede der Regionen von einem der beiden t-f-Signale gegenüber dem anderen der beiden t-f-Signale, und
• Gruppieren aller derjenigen der mehreren Regionen der Zeit-Frequenz-Ebene zu einer Gruppe, deren Regionphasendifferenz im Wesentlichen mit der geschätzten Phasendifferenz und deren Regionpegeldifferenz im Wesentlichen mit der geschätzten Pegeldifferenz übereinstimmen, wobei die Signalanteile der Gruppe als Schätzung für den Bestandteil der eigenen Stimme des Trägers dient.

According to the invention, this object is achieved by a method for estimating a component of the own voice of a wearer of a hearing device

• Placing a first microphone of the hearing device at the exit of the ear canal of an ear of the wearer or outside the auditory canal,
• Placing a second microphone of the hearing device in the ear canal, so that the second microphone is closer to the eardrum of the ear than the first microphone,
• Estimate a phase difference and a level difference of virtual microphone signals of the two microphones to each other based on a given model,
• Gaining a temporal microphone signal through each of the two microphones,
• Transforming the two temporal microphone signals to a tf signal into the time-frequency plane,
• Segment the time-frequency level into multiple regions
• Determining a region phase difference and a region level difference, respectively, for each of the regions of one of the two tf signals from the other of the two tf signals, and
• Grouping all of the plurality of time-frequency plane regions into a group whose region phase difference substantially coincides with the estimated phase difference and its region level difference substantially with the estimated level difference, the signal components of the group being estimated as the constituent of the wearer's own voice serves.

Moreover, according to the invention, a hearing device for carrying out the above method is provided, wherein the hearing device has the two microphones and a signal processing device for transforming, segmenting and grouping.

Advantageously, therefore, two microphones are placed very special. The second microphone is placed in the ear canal, while the first microphone is placed essentially at the ear canal exit or outside the ear canal (eg in the Concha or on the Pinna). Thus, the microphone arranged in the auditory canal can record significantly more portions of sound which enter the auditory canal via bone conduction than the external microphone. As a result, characteristic information based on one's own voice can be obtained. With a CASA algorithm can then own voice, so the voice of the wearer of the hearing in which the CASA algorithm is running, reliably estimate or extract.

Preferably, at least one further feature different from the phase difference and the level difference is obtained for each of the microphone signals and used for the segmentation and / or grouping. Although grouping is possible in principle only on the basis of the phase difference and the level difference is, it is beneficial to use at least one additional feature for grouping. In principle, other features may be better suited for segmentation.

Specifically, the further feature may relate to a change or a rate of change in the spectrum of the microphone signals. This has the advantage that, for example, rapid level increases (ON-SETs) can be well recognized at certain frequencies. Such signal edges are suitable for segmentation.

The further feature may also include the harmonic (degree of the acoustic periodicity) or the correlation of the two microphone signals. Harmonicity makes it easier to directly recognize language components. The correlation has the advantage that a correlate between the externally audible language and the language transmitted via bone conduction can additionally be used for the reliable determination of one's own language.

The hearing apparatus adapted to estimate a constituent of the own language according to the above principles may comprise a filter controlled by the grouping or corresponding grouping information of the signal processing means. In the filter, the regions in the time-frequency plane defined by the grouping are then used to extract corresponding signal portions, which are then presumably derived from their own voice. The segmentation and grouping technique may be used as needed, e.g. B. be repeated every time the hearing aid. This has the advantage that the filter can then be constantly adapted to the current conditions (eg seat of the hearing aid in or on the ear).

In addition, however, it is also possible to provide a hearing device which has a filter which is used for extracting the own voice of a wearer of the hearing device and filters out signal components belonging to the group of regions fall, which were obtained in advance by a method described above. The difference to the previous hearing device is therefore that the filter no longer has to be variable and thus is less expensive to produce.

The hearing device can be embodied as an in-the-ear hearing device. Alternatively, the hearing device may also be embodied as a behind-the-ear hearing device which has a hearing aid housing for carrying behind the ear and an external earpiece for carrying in the auditory canal or a sound tube for transmitting sound from the hearing device housing into the auditory canal Microphone is arranged on the external handset or the sound tube and the first microphone in the hearing aid housing. Thus, the most common types of hearing aids can benefit from the inventive method of estimating one's own voice.

FIG 1

den prinzipiellen Aufbau eines Hörgeräts gemäß dem Stand der Technik;

FIG 2

einem Querschnitt durch einen Gehörgang mit eingesetztem Hörgerät gemäß der vorliegenden Erfindung;

FIG 3

ein Blockdiagramm eines CASA-Algorithmus;

FIG 4

das Blockdiagramm von FIG 3 mit interner Struktur;

FIG 5

ein Zeit-Frequenz-Diagramm mit Nutzsignalregionen und

FIG 6

eine Schnittansicht eines Ohrs mit erfindungsgemäß gestaltetem Hinter-dem-Ohr-Hörgerät.

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

FIG. 1

the basic structure of a hearing aid according to the prior art;

FIG. 2

a cross-section through an ear canal with inserted hearing aid according to the present invention;

FIG. 3

a block diagram of a CASA algorithm;

FIG. 4

the block diagram of FIG. 3 with internal structure;

FIG. 5

a time-frequency diagram with useful signal regions and

FIG. 6

a sectional view of an ear with inventively designed behind-the-ear hearing aid.

The embodiments described in more detail below represent preferred embodiments of the present invention.

FIG. 2 schematically shows an ear canal 10 with eardrum 11, wherein an ITE hearing aid 12 is inserted into the ear canal 10. At the exit of the ear canal 10 is the outer ear 13, which is not completely drawn here. In the state inserted into the ear canal 10, the hearing device 12 has a side 14 facing the eardrum 11 and an outwardly directed side 15 facing away from the eardrum.

The hearing device 12 has a first microphone 16 on the outwardly facing side 15. This microphone 16 is only symbolically drawn outside the hearing device 12. In fact, however, the microphone is regularly in the hearing aid or at least on the surface of the hearing aid.

The first microphone 16 supplies a microphone signal m ₁ . This first microphone signal m ₁ is used for the CASA algorithm described below. However, it is also the usual signal processing device 17 of the hearing aid 12 is provided. This conventional signal processing device 17 often includes an amplifier. The output signal of the signal processing device 17 is forwarded to a loudspeaker or receiver 18 which is arranged on the side 14 of the hearing device 12 facing the eardrum 11. Also, he is symbolically located outside of the hearing aid 12 here and is mostly in the hearing aid housing.

The hearing device or the hearing device 12 has here in addition to the first microphone 16, a second microphone 19. This second microphone 19 is also located on the eardrum 11 facing side 14 of the hearing aid 12. It thus absorbs sound, which is in the room between the hearing aid 12, the eardrum 11 and the wall of the ear canal 10 results. In this often closed room especially the sound of one's own voice becomes over bone conduction entered. Among other things, the second microphone 19 picks up this sound and provides a second microphone signal m ₂ in the hearing device 12. This second microphone 19 may be referred to as an in-channel microphone.

With a in FIG. 3 symbolically represented CASA system 20, which may be integrated into the hearing aid 12, the own voice or language, ie the language of the hearing aid wearer estimated. Therefore, the CASA system 20 provides an estimate ν for a component of its own language.

In FIG. 4 is the CASA system 20 of FIG. 3 shown in detail. In the CASA system 20, the two microphone signals m ₁ and m _{2 are} supplied to an analysis unit 21. This analysis unit 21 examines each of the microphone signals m ₁ and m ₂ for specific characteristics. For this purpose, the time signals m ₁ and m _{2 are transformed} into the time-frequency range, resulting in so-called "tf signals", which can also be referred to as short-time spectra. The transformation can be performed by a high-resolution filter bank. In the analysis device 21, features are then extracted for each frequency channel of each of the two microphone signals m ₁ and m ₂ . These features are in particular the phase difference and the level difference between the two microphone signals m ₁ and m ₂ , ie in particular the phase and level difference in each point of the tf-level of the tf signals. In addition, however, further features can be extracted from the microphone signals m ₁ and m ₂ by the analysis device 21. One of these additional features may relate to so-called "on-sets". This includes, for example, rapid changes in a spectrum, which typically arise at the beginning of a vowel. Such on-sets usually represent steep edges in a tf diagram and are suitable for segmenting the tf signals.

Another feature extracted by analyzer 21 may be the harmonics, which is understood to mean the degree of acoustic periodicity. The harmonicity is often used to recognize language. As a still further feature, for example, in the analysis device 21, the correlation of the microphone signals m ₁ and m ₂ can be examined. In particular, for example, the correlation between the sound transmitted via the bone conduction into the auditory canal and the sound brought in from the outside to the ear can be analyzed. This also gives hints to your own language.

The analysis device 21 is followed on the output side by a segmentation device 22. This segmented the short-term spectra of the microphone signals m ₁ and m ₂ . This means that the segmentation device 22 calculates limits around signal components in the tf plane in such a way that according to FIG FIG. 5 Regions 24 are defined. In these regions 24 are tf signal components of a single sound source. The regions 24 in the tf plane for individual sources can be calculated in several known ways. Regions that can be assigned to a defined source therefore contain a source sound component 25. Outside of these regions 24 there are interference sound components 26 that can not be assigned to a specific source. At the time of the segmentation, however, it is not yet known which region 24 belongs to which specific source. In the FIG. 5 shown regions 24 in the tf plane are formed for both microphone signals m ₁ and m ₂ .

Subsequent to the segmentation device 22 is followed by a grouping device 23. In the grouping device 23 of a general CASA system, the segmented signal components, ie the signal components 25 in the regions 24, are organized into signal streams which are assigned to the different sound sources. In the present case, only those signal components are synthesized into a signal stream which belong to the own language of the hearing device wearer. In grouping, arbitrary regions 24 of the tf-plane can be combined with each other.

For grouping, the information of the phase difference and the level difference is used. In order to be able to use this information to determine whether a region belongs to one's own voice, the phase difference and the level difference of the two microphone signals in a model must be estimated beforehand. Based on these estimates, it can then be determined whether or not one of the segmented regions belongs to one's own vote. If the determined phase and level difference lie within a predetermined tolerance range around the estimated phase and level difference, then the region in question is counted to its own voice.

The selection of whether a region 24 is grouped with one or more multiple regions 24 is dependent on the phase difference and the level difference between the two microphone signals m ₁ and m ₂ . For grouping, however, the additional features listed above can also be used. So a group thus formed represents all parts of a short-term spectrum, which are to be joined together in order to gain only the own language or own voice from the variety of sound components. The remaining signal components in the short-term spectrum should namely be suppressed.

Now, if those regions 24 identified in the tf-level for their own language, so that a tf-filtering can be performed. For this purpose, the grouping device 23 forwards the corresponding grouping information to a filter 27 in the CASA system 20. The filter 27 is thus controlled or parameterized with the grouping information. The filter 27 receives the temporal microphone signals m ₁ and m ₂ , filters the two signals and derives an estimate for their own voice or a component ν of their own voice. In this case, the filter can use the signal components of the regions 24, unfortunately, tf signals of the two microphones or only those of one of the tf signals from a microphone to reconstruct their own voice.

From the two microphone signals m ₁ and m ₂ , which come from very specifically arranged microphones 16, 19, therefore, a special filter or special filter information is obtained by segmentation and grouping, with or from a listening situation by several sound sources is characterized, your own voice can be filtered out. As a result, a special signal model for your own language is obsolete.

The system according to the invention typically has a processing delay of a few 100 ms. This delay is necessary to extract the features and group the regions. However, this delay is not a problem in practice.

In FIG. 5 is a further embodiment with respect to the hardware structure of a hearing aid according to the invention shown. The hearing aid is a BTE hearing aid whose main component 28 is worn behind the ear, in particular behind the pinna 29. This BTE hearing aid has at the main component 28, a first microphone 30. In addition, the hearing aid here has a so-called external handset, which is mounted in the ear canal 32. Together with this external handset 31, a second microphone 33 is also mounted in the ear canal 32. Thus, the inventive extracting or estimating a component of one's own voice can also be used in a BTE hearing aid.

Thus, for the hearing device according to the invention, the CASA principle can also be used for the first time to register or extract one's own voice, since due to the special placement of the microphones, sufficient spatial information of the signals is now available. From this spatial information corresponding grouping information can be obtained, so that ultimately can be dispensed with complicated language models.

Claims (8)

A method for estimating a constituent of the own voice of a wearer of a hearing device (12) characterized by Placing a first microphone (16, 30) of the hearing device at the exit of the auditory canal (10) of an ear of the wearer or outside the auditory canal, Placing a second microphone (19, 33) of the hearing device in the auditory canal, so that the second microphone is closer to the eardrum (11) of the ear than the first microphone, Estimating a phase difference and a level difference of virtual microphone signals of the two microphones with respect to one another using a predetermined model, Each time a microphone signal (m ₁ , m ₂ ) is obtained by each of the two microphones, Transforming the two temporal microphone signals to one tf signal into the time-frequency plane, Segmenting the time-frequency level into several regions (24), Determining respectively a region phase difference and a region level difference for each of the regions of one of the two tf signals from the other of the two tf signals, and Grouping all of the plurality of regions (24) of the time-frequency plane into a group whose region phase difference substantially coincides with the estimated phase difference and its region level difference substantially with the estimated level difference, the signal portions (25) of the group being estimated as the part of the wearer's own voice. Method according to claim 1, wherein for each of the microphone signals (m ₁ , m ₂ ) at least one further feature different from the phase difference and the level difference is obtained and used for the segmentation and / or grouping. The method of claim 2, wherein the further feature relates to a change or a rate of change in the spectrum of the microphone signals (m ₁ , m ₂ ). The method of claim 2, wherein the further feature comprises the harmonic or correlation of the two microphone signals (m ₁ , m ₂ ). Hearing apparatus (12) for carrying out the method according to one of the preceding claims, wherein the hearing apparatus comprises the two microphones (16, 19, 30, 33) and a signal processing device (20) for transforming, segmenting and grouping. Hearing apparatus according to claim 5, comprising a filter (27) controlled by grouping the signal processing means. Hearing apparatus according to claim 5 or 6, which is designed as in-the-ear hearing aid. Hearing apparatus according to claim 5 or 6, which is designed as a behind-the-ear hearing aid, a hearing aid housing for carrying behind the ear and an external handset (31) for carrying in the ear canal (32) or a sound tube for transmitting sound from the Hearing aid housing in the ear canal, wherein the second microphone (33) on the external handset or the sound tube and the first microphone (30) is arranged in the hearing aid housing.

Images