EP2802158B1 - Method for adapting useful signals in binaural hearing assistance systems - Google Patents

Description

The invention relates to a method for operating a hearing aid system and a hearing aid system with at least two hearing aid devices, between which a signal path is provided, and with at least one signal processing unit which is provided for processing audio signals.

In many cases hearing impairment affects both ears, the hearing impaired should be provided binaurally with hearing aids. Modern hearing aids have signal processing algorithms which automatically vary the parameters of the hearing aids as a function of the hearing situation. In the case of binaural care, the hearing situation is assessed on both ears.

Noise and noise are omnipresent in everyday life and make speech communication more difficult, especially if there is an impairment of natural hearing. Therefore, techniques are desirable that suppress noise and noise, but change the desired sounds and tones, hereinafter also referred to as useful signals, as little as possible.

The known embodiments of techniques for suppressing noise and noise require for effective suppression of noise noises with identical as possible Nutzsignalanteilen. In practice, however, the various signals are detected by a plurality of different microphones, which are arranged on the two hearing aids. The useful signals detected by these microphones differ substantially due to ambient reverberation and the acoustic influence of the head of the wearer. The result is that downstream noise suppression techniques do not work effectively and especially when it comes to adaptive methods, they converge only very slowly or not at all. This leads to a poor suppression of the noise and additional acoustic artifacts by the noise reduction.

An adaptive method for suppressing noise is known, for example, from the publication of H. Teutsch, GW Elko "First and second-order adaptive differential microphone arrays", 7th International Workshop on Acoustic Echo and Noise Control (IWAENC), pp. 35-38, Darmstadt, Germany 2001 in which a Least Mean Square (LMS) method is used to minimize the noise.

A method for blind source separation by means of Wiener filter is from the patent application EP 2211563 A1 known.

From the US 2008/0212811 A1 a hearing aid system with two hearing aids is known in which input signals by means of a first signal channel with a first filter, and with a second signal channel with a second filter are converted into output signals. The coefficients of at least one of the filters are adjusted so that the difference between the output signals of the first and second signal channels is minimized.

The WO 2009/081193 A1 relates to a system for noise suppression, comprising a digital filter with a predetermined filter and with an adaptive filter, and with a signal processing device for generating a control signal for the adaptive filter.

The object of the present invention is to provide a method for operating a hearing aid system and a hearing aid system, by which the noise suppression in binaural hearing aid systems is improved.

This object is achieved by a method having the features of claim 1. The part of the object relating to the hearing aid system is achieved by the features of claim 12.

The method according to the invention relates to a method for operating a hearing aid system with at least two hearing aid devices. The hearing aids have a transducer for receiving an acoustic signal and conversion into a respective first audio signal. The hearing aid system further comprises a signal processing device for processing audio signals, as well as a signal connection for transmitting a first audio signal from each hearing aid device to the signal processing device. The acoustic signal has a useful signal, which arrives from a given spatial direction with respect to the hearing aid system. In the method according to the invention, the signal processing device filters the first audio signals with a filter predetermined for the given spatial direction, so that second audio signals are generated, wherein the predetermined filter is designed such that the components of the useful signal in the second audio signals are matched to one another to a greater degree than in the first audio signals; and
Filtering the second audio signals with an adaptive filter, so that third audio signals are generated, wherein the adaptive filter is designed such that the portions of the useful signal in the third audio signals are aligned to a higher degree than in the second audio signals.

In the first step, the method according to the invention provides second audio signals in which the components of the useful signal are matched to one another to a greater degree than in the first audio signals. Advantageously, therefore, the adaptive filter of the second step converges much faster and therefore provides third audio signals in which the alignment of the useful signals is further improved. This allows subsequent noise suppression methods from these third audio signals to form better reference signals for the useful and interfering signals and noise almost completely to suppress. By preprocessing the first audio signals in the predetermined filter, the method of the invention allows shorter adaptive filters with less computational power requirement and faster convergence of the adaptive filter, which advantageously allows adaptation to rapidly changing environmental conditions.

The hearing aid system according to the invention for carrying out the method according to the invention shares its advantages.

Advantageous developments of the method and the hearing aid system are specified in the subclaims.

Thus, in a preferred embodiment of the invention, the signal processing device has means for dividing the first audio signals into first audio signals in predetermined frequency ranges, wherein the steps of the method according to the invention are performed separately for each audio signal in a frequency range.

This makes it possible to optimize the predetermined filter and the adaptive filter for the respective frequency range as well as to integrate the method in the architecture of a hearing aid with a frequency response correction corresponding to the impairment of the patient.

In one possible embodiment, the predetermined filter is defined by a predetermined number of predetermined coefficients.

A filter with a predetermined coefficient number is easy to implement and advantageously requires only a small memory space.

In one embodiment, it is conceivable that the predetermined number of the predetermined coefficients is different in different frequency ranges.

Thus, the need for computing power for the predetermined filter can be optimized in an advantageous manner by more accurate filtering takes place only in frequency ranges with corresponding noise. In addition, the memory requirement for the filter is further reduced.

In a preferred embodiment, predetermined coefficients for the predetermined filter are predetermined for at least two predetermined spatial directions.

It is therefore easily possible to switch the predetermined filter from one direction in which the filtering is particularly effective to another direction, for example, if the source for the useful signal changes the location.

In a preferred embodiment of the method according to the invention, this further comprises a step of estimating a spatial direction of the useful signal, wherein in the step of filtering the first audio signals, the predetermined filter is predetermined by the coefficients whose assigned predetermined spatial direction is closest to the estimated spatial direction.

Thus, advantageously, the predetermined filter can be set to different predetermined spatial directions, wherein in each case the selected spatial direction of the spatial direction of the source of the useful signal is closest, so that the predetermined filter achieves an optimal filtering effect.

In one possible embodiment of the method, a reference signal for a noise is derived from the second and / or third audio signals by subtraction.

It is advantageous that in the second audio signals by the predetermined filter or in the third audio signals by the adaptive filter already the proportions of the useful signal to a higher degree aligned with each other are so that they almost completely cancel each other by the difference and a clearer reference signal for the noise is provided. For example, subsequent noise suppression techniques using these reference signals produce less artifacts and a cleaner payload.

In one possible embodiment of the method, a reference signal for a useful signal is derived from the second and / or third audio signals by summation.

It is advantageous that in the second audio signals by the predetermined filter or in the third audio signals by the adaptive filter already the proportions of the useful signal are aligned to a higher degree to each other, so when adding a signal with a better signal-to-noise ratio available stands.

In one possible embodiment of the method according to the invention, the reference signal is used for a useful signal for detecting a voice activity.

Because of the better signal-to-noise ratio, the reference signal for the useful signal, for example, when used to detect a voice activity (VAD, voice activity detection) allows a safer detection.

In a preferred embodiment of the method according to the invention, the predetermined filter is a finite impulse response (FIR) filter.

An FIR filter advantageously has no tendency to oscillate and thus ensures a convergence of a provided signal under all circumstances.

The above-described characteristics, features, and advantages of this invention, as well as the manner in which they are achieved, become clearer and more clearly understandable with the following description of the embodiments, which are explained in more detail in connection with the drawings.

Fig. 1

eine schematische Darstellung eines erfindungsgemäßen Hörhilfesystems;

Fig. 2

ein Ablaufdiagramm eines erfindungsgemäßen Verfahrens;

Fig. 3

eine schematische Darstellung der Funktionsaufteilung einer Ausführungsform des binauralen Hörhilfesystems;

Fig. 4

eine schematische Darstellung für einen Testaufbau zur Bestimmung der richtungsabhängigen vorbestimmten Filterkoeffizienten.

Show it:

Fig. 1

a schematic representation of a hearing aid system according to the invention;

Fig. 2

a flow diagram of a method according to the invention;

Fig. 3

a schematic representation of the functional distribution of an embodiment of the binaural hearing aid system;

Fig. 4

a schematic representation of a test setup for determining the direction-dependent predetermined filter coefficients.

Fig. 1 shows the basic structure of a hearing aid system 100 according to the invention. The hearing aid system 100 has two hearing aid devices 110, 110 '. In a hearing aid housing 1, 1 'for carrying behind the ear, one or more microphones 2, 2' for receiving the sound or acoustic signals from the environment are installed. The microphones 2, 2 'are transducers 2, 2' for converting the sound into first audio signals. A signal processing unit 3, 3 ', which is also integrated in the hearing aid housing 1, 1', processes the first audio signals. The output signal of the signal processing unit 3, 3 'is transmitted to a loudspeaker or receiver 4, 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 of the signal processing unit 3, 3 'takes place by means of a likewise integrated into the hearing aid housing 1, 1' battery 5, 5 '.

Furthermore, the hearing aid system 100 has a signal connection 6, which is designed to transmit a first acoustic signal from the signal processing device 3 to the signal processing device 3 '. It is provided in the preferred embodiment that also signal processing device 3 'transmits a first acoustic signal to the signal processing device 3 in the opposite direction. Furthermore, it is conceivable to transmit the signals of several or all of the microphones 2, 2 'in each case to the other hearing aid device 110, 110'.

As a signal connection 6 are wired, optical or wireless connections such. Bluetooth conceivable.

Fig. 2 shows a schematic flow diagram of a method according to the invention in the signal processing device 3, 3 '.

In step S10, a spatial direction of a desired signal is estimated by the signal processing device 3, 3 '. This can be done for example by time or phase differences for a prominent signal in the first audio signals. An analysis of the amplitude of this distinctive signal in the first audio signals is also conceivable. From the estimated spatial direction of the useful signal, the signal device selects a spatial direction from a set of predetermined spatial directions which is closest to the estimated spatial direction of the useful signal. If no distinctive signal is present, an estimate of the spatial direction is used at an earlier time or a predetermined direction, e.g. selected in a straight line in front of the face of the wearer of the hearing aid system. It is also conceivable to undertake a new estimation of the spatial direction only in the case of a multiple occurrence of a distinctive signal having the same characteristics.

In step S20, the first audio signals are filtered by the predetermined filter 31. In this case, the predetermined filter 31 uses predetermined filter coefficients for the given one Spatial direction of the wanted signal. For example, for each element of the set of given spatial directions, a set of predetermined filter coefficients for the predetermined filter 31 may be stored in a look-up table. As a result of step S20, the predetermined filter 31 provides second audio signals. In these second audio signals, the useful signal is already equalized as a signal component by the use of the predetermined filter 31, ie, the differences in amplitude and phase of the useful signal are smaller between the individual second audio signals than between the corresponding first ones Audio signals. Under real ambient conditions, a compensation in the range of 5 - 10 dB can be achieved.
In step S30, the adaptive filter 32 filters the second audio signals. As an adaptive filter, for example, the Least Mean Square (LMS) method described by Teutsch and Elko can be used to minimize the noise. Due to the pre-filtering by the predetermined filter 31, the filter effect by the adaptive filter 32 may be smaller. Therefore, for example, a filter with fewer coefficients can be used, and therefore an algorithm for adaptively adapting the filter coefficients to the second audio signals converges faster.

As a result of step S30, the adaptive filter 32 provides third audio signals. In these third audio signals, the useful signal is already adjusted by the application of the adaptive filter 32 as a signal component to an increased extent, that is, the differences in amplitude and phase of the useful signal between the individual third audio signals are even lower than between the corresponding second audio signals.

In a conceivable embodiment, a reference signal for a noise can be formed in a conceivable embodiment by subtracting third audio signals, which is used in a step S50 by a noise suppression method is to suppress noise in the audio signals before they are output via the handset 4 as acoustic signals. Due to the filtering, the useful signal in the different parts of the third audio signal is almost identical, so that it almost completely extinguishes during difference formation. Therefore, artefacts which distort the wanted signal hardly occur in the subsequent noise suppression method.

It is also conceivable that in a step S60 by summation of third audio signals, a reference signal for the useful signal is generated. Since the components of the useful signal are matched by the filtering, but the disturbances are statistically distributed, addition of the useful signal to the interference is further emphasized. Therefore, the reference signal for the wanted signal obtained in step S60 has a better signal-to-noise ratio and can be used for voice actvity detection in a further step S70.

The steps S10 to S70 may be applied to audio signals having a single continuous frequency range. In a preferred embodiment, however, the first audio signals are first divided into different frequency ranges before all or individual steps S10 to S70 are applied to the individual frequency ranges. This has the advantage of using a different gain and other signal processing, such as a dynamic compression in a hearing aid to compensate for a hearing loss of the patient for individual frequency ranges. Means for dividing the audio signals into frequency ranges are therefore usually already present and the steps S10 to S70 can be easily incorporated into the existing processing process. However, it is also conceivable that individual steps, such as, for example, step S10 for estimating a given spatial direction, take place jointly for all frequencies.

Fig. 3 shows a corresponding schematic representation of a hearing aid system 100 according to the invention with a frequency distribution.

On the left is in Fig. 3 the head 200 of a patient is shown with an array of two microphones 2, 2 'on each side of the head. The other functional blocks are then no longer specified in their relative position to the head.

The first audio signals of the microphones 2, 2 'are supplied via a signal connection 6 to both hearing aid devices 110, 110' and the signal processing device 3, 3 '. In an analysis filter bank 10, 10 'realized in the signal processing units 3, 3', each of the four first audio signals in each hearing aid device 110, 110 'is divided into a plurality of frequency bands. For the individual frequency ranges of the first audio signals, the signal processing device 3, 3 'carries out the step S20, which is represented by the function block of the predetermined filter 31, 31'. The second audio signals are supplied to the adaptive filter 32, 32 ', which performs step S30 on the individual frequency ranges of the second audio signals and provides third audio signals. Before conversion into acoustic signals, which are output via the earphones 4, 4 ', synthesis of the individual frequency ranges of the third audio signals takes place in the synthesis filter bank 33, 33', so that four third audio signals each comprise a single, at least parts of the audible range Frequency range arise.

The analysis filter bank 10, 10 ', which divides the first audio signals into frequency ranges, can be realized for example by a short-time Fourier transformation. The further processing of the second and third audio signals then takes place in the short-time Fourier range, before they are in the synthesis filter bank 33, 33 'in turn assembled by, for example, an inverse short-term Fourier transform to form an audio signal.

In the Fig. 3 shown division of the function blocks is only an example. Thus, it would also be conceivable that the signal connection 6 is not connected upstream of the signal processing unit 3, 3 ', but the signal processing unit 3, 3' itself provides protocols, signaling and interfaces, so that only the physical connection level of the signal connection 6 outside the signal processing unit 3, 3 ' takes place.

It would also be possible for only one hearing aid device 110, 110 'of the hearing aid system 100 to have the signal processing unit 3 configured to carry out steps S10 to S70, while the other signal processing unit 3' is for providing the first audio signals and transmitting the audio signals via the signal connection 6 limited. For example, the energy consumption for a parallel processing of all signals in both hearing aids 110, 110 'could be avoided. Other architectures, for example, with a central, remote signal processing unit 3 are conceivable.

Fig. 4 shows a schematic representation of a test setup for determining the direction-dependent predetermined filter coefficients. The filters 31 are predetermined, for example by predetermined filter coefficients. These filter coefficients can either be theoretically determined by a model calculation, or determined by measurements and adaptive methods in the laboratory and stored in tables in the signal processing unit 3, 3 '. In Fig. 4 an artificial head 200 (eg KEMAR) is shown on which microphones 2, 2 'are arranged. A measuring device, not shown, receives the audio signals of the microphones as a function of the spatial direction of a signal source 120 given by the angle Φ, which emits a useful signal as sound. Modeling and optimization methods can then be used to determine a set of coefficients that are suitable for predetermined spatial directions under the idealized conditions the measurement equalizes the components of the useful signal in the second audio signals. In the hearing aid system 100, a spatial direction of a real useful signal source can later be estimated via the step S10, and in the predetermined filter 31 the corresponding coefficient set for a predetermined spatial direction closest to the estimated spatial direction can be selected and used for filtering. Since the real conditions deviate from the ideal conditions of the measurement, the matching of the components of the useful signal in the second audio signals is not completely achieved, which is why the method according to the invention provides for a subsequent adaptive filter 32, 32 '.

Claims (12)

1. Method for operating a hearing aid system (100) having at least two hearing aid devices (110, 110'), wherein the hearing aid devices (110, 110') have a transducer (2, 2') for picking up an audible signal and converting it into a respective first audio signal, wherein the hearing aid system (110, 110') has a signal processing apparatus (3, 3') for processing audio signals and the hearing aid system has a signal connection (6) for transmitting the respective first audio signal from each hearing aid device (110, 110') to the signal processing apparatus (3, 3'), and wherein the signal processing apparatus (3, 3') has a predetermined filter (31, 31') having predetermined filter coefficients for a given spatial direction,
   characterized
   in that the signal processing apparatus (3, 3') carries out the following steps for aligning components of a useful signal in the picked-up audio signals of the audible signal in third audio signals, wherein the useful signal arrives from the given spatial direction in relation to the hearing aid system:

   filtering (S20) the first audio signals using a filter (31, 31') predetermined for the given spatial direction, so that second audio signals are generated, wherein the predetermined filter (31, 31') is designed such that the components of the useful signal are aligned with one another to a greater extent in the second audio signals than in the first audio signals; and

   filtering (S30) the second audio signals using an adaptive filter (32, 32'), so that third audio signals are generated, wherein the adaptive filter (32, 32') is designed such that the components of the useful signal are aligned with one another to a greater extent in the third audio signals than in the second audio signals.
2. Method according to Claim 1, wherein the signal processing apparatus has means (10, 10') for splitting the first audio signals into first audio signals in prescribed frequency ranges, and the steps of Claim 1 are carried out separately for each audio signal in a frequency range.
3. Method according to Claim 1 or 2, wherein the predetermined filter (31, 31') is defined by a predetermined number of predetermined coefficients.
4. Method according to Claim 2 and 3, wherein the predetermined number of predetermined coefficients is different in different frequency ranges.
5. Method according to one of the preceding claims, wherein predetermined coefficients for the predetermined filter are predetermined for at least two predetermined spatial directions.
6. Method according to Claim 5, wherein the method furthermore has a step for estimating a spatial direction of the useful signal, and the step of filtering the first audio signals involves the predetermined filter (31, 31') being predetermined by the coefficients whose associated predetermined spatial direction is closest to the estimated spatial direction.
7. Method according to one of the preceding claims, wherein a reference signal for a noise is derived from the second and/or third audio signals by means of difference formation.
8. Method according to one of the preceding claims, wherein a reference signal for a useful signal is derived from the second and/or third audio signals by means of summation.
9. Method according to Claim 8, wherein the reference signal for a useful signal is used to detect a voice activity.
10. Method according to one of the preceding claims, wherein the predetermined filter (31, 31') is a finite impulse response filter.
11. Method according to one of the preceding claims, wherein the adaptive filter (32, 32') uses a least mean square (LMS) method to minimize the noise.
12. Hearing aid system (100) having at least two hearing aid devices (110, 110'), wherein the hearing aid devices (110, 110') have a transducer (2, 2') for picking up an audible signal and converting it into a respective first audio signal, wherein the hearing aid system has a signal processing apparatus (3, 3') for processing audio signals and the hearing aid system has a signal connection (6) for transmitting the respective first audio signal from each hearing aid device (110, 110') to the signal processing apparatus (3, 3'), wherein the signal processing apparatus (3, 3') is configured to carry out the method according to one of Claims 1 to 11.

Images