EP2120484B1 - Method for operating a hearing device and hearing device - Google Patents

Description

The invention relates to a specified in claim 1 method for operating a hearing aid and a specified in claim 6 hearing aid.

In a conversation between people noise or unwanted acoustic signals are omnipresent. These interfere with the human voice of a person or with a desired acoustic signal. Hearing aid carriers are particularly susceptible to noise and unwanted acoustic signals. Talking in the background, audible interference with electronic devices such as cell phones, and ambient noise or sounds may make it difficult for a person with a hearing aid to understand a desired speaker. Reducing the level of noise in an acoustic signal coupled with an automatic focus on a desired acoustic signal component can significantly improve the performance of a digital speech processor as used in modern hearing aids.

Hearing aids with digital signal processing include one or more microphones, A / D converters, digital signal processors and speakers. As a rule, digital signal processors divide the incoming signals into a plurality of frequency bands. Within each band, signal amplification and processing may be individually adjusted in accordance with the requirements of a particular wearer of the hearing aid. Furthermore, algorithms for feedback and noise minimization are available in digital signal processing, but they also have disadvantages. A disadvantage of the currently existing algorithms for noise minimization is, for example, their limited improvement in hearing aid acoustics when speech and background noise are in the same frequency region and they are therefore unable to distinguish between spoken language and background noise. In the field of acoustic signal processing this is one of the most frequent tasks, namely to filter out different, overlapping, acoustic signals of one or a plurality of them. This is also referred to as the so-called "cocktail party problem". Here, the most diverse sounds, such as music and conversations mix to an indefinable soundscape. Nevertheless, it is generally not difficult for a person without hearing loss to talk to a conversation partner in such a situation. It is therefore desirable for hearing aid users to be able to talk in just as situations, as people without hearing loss.

There are in the acoustic signal processing spatial, z. B. directional microphone or beamforming, statistical, z. B. Blind Source Separation (BSS) or mixed methods, which may include: a. by means of algorithms from several simultaneously active sound sources can separate a single or a plurality thereof. Thus, BSS allows statistical signal processing of at least two microphone signals to perform a separation of source signals without prior knowledge of their geometric arrangement. This method has advantages over conventional directional microphone solutions when used in hearing aids. By principle, with a BSS method with n microphones, up to n sources can be separated, i. H. n Generate output signals.

Methods for BSS are well known in the literature, with sound sources analyzed by analyzing at least two microphone signals. A good overview is the post-published patent DE 10 2006 047 982 ,

The directional microphone control in the sense of a BSS is subject to ambiguity as soon as several competing sources of use, eg. B. speaker, present simultaneously. The BSS allowed in principle, the separation of the different sources, if they are spatially separated. The ambiguity, however, reduces the potential benefit of a directional microphone, although it is precisely in such scenarios that a directional microphone can be very useful for improving speech intelligibility.

The hearing aid or the mathematical algorithms for BSS are in principle faced with the problem of having to decide which of the signals generated by the BSS should be passed on most advantageously to the hearing aid wearer. This is a problem that in principle is impossible for the hearing aid since the selection of the desired acoustic source directly depends on the momentary will of the hearing aid wearer and thus can not be present as an input variable for a selection algorithm. The choice made by this algorithm must therefore be based on assumptions about the probable will of the listener.

In the prior art, a preference is given to an acoustic signal by the hearing aid wearer from a 0 ° direction, that is to say the viewing direction of the hearing aid wearer. This is realistic in that, in an acoustically difficult situation, the hearing aid wearer would look at his current interlocutor in order to obtain further indications which increase the speech intelligibility of the interlocutor (eg lip movements). However, this forces the hearing aid wearer to look at his interlocutor so that the directional microphone leads to increased speech intelligibility. This is especially troublesome when the hearing aid wearer wants to talk to just one person, ie is not involved in communication with multiple speakers, and does not always want / need to view his interlocutor.

Out EP 1 912 472 A1 a method for operating a hearing aid and a hearing aid is known. From a recorded ambient sound electrical acoustic signals are generated from which a signal processing an electrical voice signal with a high probability of speech is identified and selected. The electrical speech signal is selectively taken into account in an output sound of the hearing aid so that it at least emerges acoustically for the hearing aid wearer in comparison with another acoustic source and is thus better perceived by the hearing aid wearer.

In EP 1 912 474 A1 discloses a method for operating a hearing aid and a hearing aid. From a recorded ambient sound electrical acoustic signals are generated, from which by signal processing, an electrical speaker signal is identified and selected by a database with voice profiles preferred speakers. The electrical speaker signal is selectively taken into account in an output sound of the hearing aid so that it at least emerges acoustically for the hearing aid wearer in comparison with another acoustic source and is thus better perceived by the hearing aid wearer.

It is therefore an object of the invention to provide an improved method for operating a hearing device, as well as an improved hearing aid, with which it can be decided which Output signals of a source separation, in particular a BSS, the hearing aid wearer are supplied acoustically.

According to the invention, the stated object is achieved by the method of independent claim 1 and the hearing aid of independent claim 6.

The invention comprises a method for operating a hearing device, wherein electrical acoustic signals are generated by the hearing device from a recorded ambient sound. These are weighted according to their degree of membership to a predeterminable acoustic signal class and mixed together to form an output sound signal. The weight of the acoustic signal is greater or smaller, the higher the degree of membership. The advantage of this is that a hearing aid user a desired signal from a variety of ambient sound signals can be presented.

In a further development, the degree of membership can be determined by the characteristics volume, frequency range, basic speech frequency, cepstral coefficients and / or temporal course of the acoustic signals. This achieves high flexibility.

In a further embodiment, the predeterminable acoustic signal class may include the classes voice or human voice, in a predeterminable frequency band, male voice, female voice, child's voice, voice of a predefinable person, music and ambient noise. This offers the advantage of a wide choice for a hearing device user.

The predeterminable acoustic signal class may also comprise any combination of the classes.

The electrical acoustic signals are generated from the ambient sound by means of a blind-source separation method. This results in a good sound signal separation.

The degree of membership is determined by a feature analysis of the electrical acoustic signals, wherein a probability of belonging to a predeterminable acoustic signal class is determined for the electrical acoustic signals. The advantage of this is the simple mathematical weighting.

According to the invention, a hearing aid is also provided with at least one microphone for recording ambient sound and with a demixing unit for generating electrical acoustic signals from the recorded ambient sound. The hearing device comprises a signal processing unit, by the acoustic signals which are weighted according to their degree of membership to a predetermined acoustic signal class and mixed together to form an output sound signal, wherein the weight of the acoustic signal is greater or smaller, the higher the degree of affiliation. This makes it possible to switch between acoustic signal classes "soft".

In a further development, the demixing unit may comprise a blind-source separation module.

In a further embodiment, the signal processing unit may comprise at least one classification module, at least one weight determination module, at least one multiplier and at least one adder.

Furthermore, the hearing device can comprise an acoustic signal class input unit with which the desired, specifiable acoustic signal class is transmitted to the hearing aid. This can be arranged on the hearing aid or in a remote control.

Other features and advantages of the invention will become apparent from the following explanations of several embodiments with reference to schematic drawings.

Figur 1:

ein Blockschaltbild eines Hörgeräts mit Blind Source Separation nach dem Stand der Technik und

Figur 2:

ein Blockschaltbild eines erfindungsgemäßen Hörgeräts.

Show it:

FIG. 1:

a block diagram of a hearing aid with blind-source separation according to the prior art and

FIG. 2:

a block diagram of a hearing aid according to the invention.

The FIG. 1 shows the state of the art of a hearing aid 1 with three microphones 2 and a demixing unit 5 according to the blind source separation method. Three signal sources generate three ambient acoustic sound signals s1, s2, s3, which are received by the three microphones 2 and converted into electrical microphone signals X1, X2, X3. The three microphone signals x1, x2, x3 are each supplied to a signal input of the demixing unit 5. In the demixing unit 5, the blind-source separation process takes place with the aid of which the ambient sound signals s1, s2, s3 can be reconstructed from the mixed electrical microphone signals x1, x2, x3. Thus, at three outputs of the demixing unit 5, three electrical acoustic signals s1 ', s2', s3 'are available.

In the simplest case, a hearing device user can now select between the three separately reproduced acoustic signals s1 ', s2', s3 'by means of a selection switch 7 in a post-processor module 6. In FIG. 2 the electrical acoustic signal s2 'was selected and passed on to a handset 3. The demixing unit 5 and the post-processor module form a signal processing unit 4.

The listener 3 sends as an acoustic output signal s2 '', which corresponds approximately to the ambient acoustic signal s2. With the help of the hearing aid 1 in FIG. 1 Thus, different acoustic input signals can be separated and output separately according to the preferences of a hearing device user via the handset 3.

A hearing aid user does not always want such a hard switching between different input signal sources.

Also, it is not always possible for a demixing unit 5 to process the signals in such a clean and secure manner. An improved presentation of different ambient sound signals therefore provides the device according to the invention FIG. 2 ,

FIG. 2 shows a hearing aid 1 with three microphones 2, a signal processing unit 4 and a receiver or loudspeaker 3. Three ambient sound signals s1, s2, s3 are received by the microphones 2 and forwarded as microphone signals x1, x2, x3 to the signal processing unit 4. The processed by the signal processing unit 4 microphone signals x1, x2, x3 are then forwarded to an input of the handset 3 and presented as an acoustic output sound signal s the hearing device user.

In the signal processing unit 4, the microphone signals x1, x2, x3 are processed with the aid of a demixing unit 5 and passed as demixed electrical acoustic signals s1 ', s2', s3 'to the further processing units. On the one hand, the electrical acoustic signals s1 ', s2', s3 'reach inputs of multipliers 10, on the other hand inputs of a classification module 8. With the aid of an acoustic signal class input unit 12, a hearing device user can specify a preferred acoustic signal class. This specification is passed to the classification module 8 and processed in this. The preselected acoustic signal class may include, for example, a male voice, a female voice, a child's voice or even a certain frequency range, or generally human voice, or music, etc. The classification module 8 calculates the probability with which an electrical acoustic signal s1 ', s2', s3 'belongs to a specific acoustic signal class. With the help of a weight determination module 9, this degree of affiliation is now weighted accordingly. For this purpose, the degrees of membership of the classified signals are routed from outputs of the classification module 8 to inputs of the weighting module 9. The weight determination module 9 now determines the weights g1, g2, g3 such that the higher the degree of belonging to the preselected class has been determined, the higher the weight of an acoustic signal is selected. The weights g1, g2, g3 are routed to these respective inputs of the multipliers 10. In the multipliers 10, the electrical acoustic signals s1 ', s2', s3 'are now multiplied by the weights g1, g2, g3. From outputs of the multipliers 10, the weighted electrical acoustic signals are passed to an adder 11. In the adder 11, these signals are added and provided to the output of the adder 11. Subsequently, the electrical signal at the output of the adder in the receiver 3 is converted into an output sound signal S.

LIST OF REFERENCE NUMBERS

1

hearing Aid

2

microphone

3

Handset / speaker

4

Signal processing unit

5

unscrambling

6

Post-processor module

7

selector switch

8th

classification module

9

Weight determination module

10

multipliers

11

adder

12

Acoustic signal class input unit

g1, g2, g3

Weight factors / weights

s

Output sound signal

s1, s2, s3

Ambient sound signals

s1 ', s2', s3 '

electrical acoustic signals

x1, x2, x3

microphone signals

Claims (8)

1. Method for operating a hearing aid (1), wherein the hearing aid (1) takes a picked-up ambient sound (s1, s2, s3) and produces electrical acoustic signals (s1', s2', s3') by means of a blind source separation method, wherein the electrical acoustic signals (s1', s2', s3') each have a probability of association with a prescribable acoustic signal class ascertained for them as a degree of association by dint of feature analysis of the electrical acoustic signals (s1', s2', s3'), and wherein the electrical acoustic signals (s1', s2', s3') are weighted in accordance with their respective degree of association with the prescribable acoustic signal class and are mixed together to form an output sound signal (s), the respective weight (g1, g2, g3) of the acoustic signal (s1', s2', s3') being all the greater or all the smaller the higher the respective degree of association.
2. Method according to Claim 1,
   characterized in that the degree of association is determined by the features volume, frequency range, fundamental voice frequency, cepstral coefficients and/or time characteristic of the acoustic signals (s1', s2', s3').
3. Method according to Claim 1 or 2,
   characterized in that the prescribable acoustic signal class comprises the following classes:

   - speech or human voice,

   - in a prescribable frequency band,

   - man's voice, woman's voice, child's voice,

   - voice of a prescribable person,

   - music and

   - ambient noise.
4. Method according to Claim 3,
   characterized in that the prescribable acoustic signal class comprises any combination of the classes.
5. Computer program product having a computer program that has software means for performing a method according to one of Claims 1 to 4 when the computer program is executed in a control unit.
6. Hearing aid (1) having at least one microphone (2) for picking up an ambient sound (s1, s2, s3) and having a demixing unit (5) for producing electrical acoustic signals (s1', s2', s3') from the picked-up ambient sound (s1, s2, s3), wherein the demixing unit (5) comprises a blind source separation module and a signal processing unit (4) that acoustic signals (s1', s2', s3') that can be weighted in accordance with their respective degree of association with the prescribable acoustic signal class and can be mixed together to form an output sound signal (s), the respective weight (g1, g2, g3) of the acoustic signal (s1', s2', s3') being all the greater or all the smaller the higher the respective degree of association, which can be determined by dint of feature analysis of the electrical acoustic signals (s1', s2', s3'), wherein the electrical acoustic signals (s1', s2', s3') can have a probability of association with a prescribable acoustic signal class ascertained for them as a respective degree of association.
7. Hearing aid (1) according to Claim 6,
   characterized in that the signal processing unit (4) comprises at least one classification module (8), at least one weight ascertainment module (9), at least one multiplier (10) and at least one adder (11).
8. Hearing aid (1) according to Claim 6 or 7,
   characterized by an acoustic signal class input unit (12) that can be used to transmit the desired, prescribable acoustic signal class to the hearing aid (1).

Images