The invention relates to a bilateral hearing aid system comprising at least one bone conduction output transducer.

Examples of bone conduction hearing aid systems are described in US 2009/0245553 Al and US 2009/0247810 Al.

Bone conduction hearing aids are used by patients who cannot benefit from electro- acoustic hearing aids. Most of them are suffering from malformed ears, conductive hearing loss or single-sided deafness.

In general, bone conduction hearing aids use a mechanical transducer coupled to the skull to directly transfer sound vibrations through the bone to the cochlea, thereby bypassing the outer and the middle ear.

In case of non- implanted devices the transducer may be incorporated in a BTE (Behind The Ear) housing or an ITE (In the Ear) shell, having direct contact to the skull with the skin in- between, or it may be coupled to the skull using a head belt or an eyeglass adapter, or it may be coupled to the teeth.

In bone- anchored devices, surgically implanted abutments in the skull are used to achieve an improved coupling between the transducer and the skull. Such abutments may be magnets offering a strong transcutaneous magnetic coupling with the externally located transducer, or they may be designed as a percutaneous "screw" on which the transducer is sitting.

Usually the transducer forms part or is connected to an external sound processor, which typically is a BTE- or ITE-type device comprising one or more microphones, a signal processing and amplification unit and a driver for the transducer. The sound processor device is usually placed close to the ear to provide the most natural sound pick up position for the microphones. The transducer may be integrated in the sound processor housing or it may be a separate element connected by wire or by a wireless radio link to the sound processor.

It is generally desirable to fit hearing aids bilaterally in order to achieve the well-known advantages of binaural hearing in terms of speech understanding, sound quality and spatial hearing. However, the benefit of bilateral fittings is limited in case of bone conduction hearing aids. The reason is that the interaural time differences (ITD) and interaural leveled differences (ILD) cues are disturbed due to the strong transcranial cross-talk effect of bone conduction. Bone-conducted vibrations reach the contralateral cochlea with an average attenuation of just about 10 dB compared to the ipsilateral cochlea. By contrast, for air conduction, i.e. electro- acoustic, hearing aids, the interaural attenuation typically is more than 50 dB. Hence, in case of bone conduction there is an unnatural interference of the sound coming from the ipsilateral transducer and the contralateral transducer. The result are deteriorated ITDs and ILDs, so that the benefit of binaural hearing is quite small compared to what could be expected is the cochleae received proper stimuli.

WO 2009/101622 A2 relates to a sound system for reproducing recorded sound, comprising several loudspeakers and bone conduction speakers to be located at the right side and the left side of a user's head. It is mentioned that transcranial cross talking occurs with the use of bone conduction speakers, and a theoretical analysis of this effect is described. It is also mentioned that interesting effects can be achieved by controlling such cross-talking effect.

The article "Head-related two-channel stereophony with loudspeaker reproduction", by P. Damaske, JASA, VoI 50, 1971 relates to cross- talk compensation techniques for virtual acoustic imaging with two free- field loudspeakers.

It is an object of the invention to provide for a bilateral hearing aid system comprising at least one bone conduction output transducer, wherein binaural hearing effects should be preserved as far as possible. It is a further object of the invention to provide for a corresponding hearing assistance method.

According to the invention, these objects are achieved by a bilateral bone conduction hearing aid system as defined in claim 1 and a corresponding hearing assistance method as defined in claim 10, and by a bimodal hearing aid system as defined in claim 9 and a corresponding hearing assistance method as defined in claim 21, respectively.

The invention is beneficial in that, by exchanging cross-talk compensation signals generated according to the respective estimated transcranial transfer function between the right ear side and the left ear side and by subjecting such contralateral cross-talk compensation signal from the "direct" ipsilateral signal prior t> supplying the ipsilateral signal as input to the bone conduction output transducer, cross talk compensation can be achieved, thereby preserving binaural effects.

Preferred embodiments of the invention are defined in the dependent claims.

Hereinafter examples of the invention will be illustrated by reference to the attached drawings, wherein:

Fig. 1 is a block diagram of an example of a hearing aid system according to the invention comprising two bone conduction output transducers;

Fig. 2 is a signal processing model of an example of a fitting of a hearing aid system according to the invention;

Fig. 3 is a block diagram of an example of an estimation of the transfer functions used in a hearing aid system according to the invention; and

Fig. 4 is a block diagram of an example of a hearing aid system according to the invention comprising one bone conduction output transducer and one electro- acoustic or electro- mechanical outpit transducer.

Fig. 1 shows a block diagram of an example of a bone conduction hearing aid system according to the invention, comprising a right ear hearing aid 1OA and a left ear hearing aid 1OB. The right ear hearing aid 1OA comprises a microphone arrangement 12A for capturing audio signals from ambient sound, an audio signal processing unit 14A for processing the audio signals captured by the microphone arrangement 12A and a bone conduction output transducer 16A. The right ear hearing aid 1OA also comprises a filter unit 18A for generating a right ear cross- talk compensation signal from the processed audio signals of the right ear audio signal processing unit 14A, according to an estimated transcranial transfer function from the right ear bone conduction output transducer 16A to the left ear cochlea 20B and an adder unit 22A for adding a left ear cross talk compensation signal received from the left ear hearing aid 1OB to the processed audio signals produced by the right ear audio signal processing unit 14A. The units 14A, 18A and 22A typically will be implemented by a digital signal processor (DSP) 24A. The combined output signal of the adder 22A forms a right ear output audio signal, which is supplied, after having undergone amplification in a power amplifier 26A, as input to the output transducer 16A, which is located at or close to the user's right ear, and hence close to the user's right ear cochlea 2OA, in order to stimulate the right ear cochlea 2OA according to the right ear audio output signals.

The output transducer 16A may be a bone conduction transducer of any type. In particular, the output transducer 16A may be for direct contact with the skin at the user's skull, or it may be for engagement with implantable abutments. The output transducer 16A also may be coupled to the skull using a head belt or an eyeglass adapter, or it may be coupled to the teeth. The microphone arrangement 12A may comprise a single microphone or a plurality of spaced- apart microphones for enabling acoustic beam forming.

The right ear hearing aid 1OA may be realized as a BTE hearing aid, ITE hearing aid or as part of an eyeglass frame. The output transducer 16A may be integrated in the housing of the hearing aid 1OA, or it may be realized as an external part connected by wire or by using a wireless radio link to the hearing aid 1OA.

The left ear hearing aid 1OB comprises the like components as the right ear hearing aid 1OA, but in a mirror- like manner, i.e. the left ear filter unit 18B is for generating a left ear cross- talk compensation signal from the processed audio signals of the left ear signal processing unit 14B according to an estimated transcranial transfer function from the left ear bone conduction output transducer 16B to the right ear cochlea 20A, and the adder unit 22B is for adding the right ear cross- talk compensation signal generated by the right ear filter unit 18B to the processed audio signals produced by the left ear audio signal processing unit 14B.

The hearing aids 1OA, 1OB also include means for exchanging the cross-talk compensation signals between the hearing aids, i.e. means for sending the right ear cross-talk compensation signal from the right ear filter unit 18A to the left hear hearing aid 1OB and for sending the left ear cross- talk compensation signal from the left ear filter unit 18B to the right ear hearing aid 1OA. Such signal exchange may be realized by a wire connection indicated at 28 A and 28B in Fig. 1. Alternatively, the hearing aids 1OA, 1OB may comprise means for establishing a bidirectional wireless link 35 between the right ear hearing aid 1OA and the left ear hearing aid 1OB, which means include a right ear transceiver 3OA of the right ear hearing aid 1OA and a left ear transceiver 3OB in the left ear hearing aid 1OB, as well as respective antennas 32A in the right ear hearing aid 1OA and 32B in the left ear hearing aid 1OB.

Such wired or wireless bidirectional audio link between the right ear hearing aid 1OA and the left ear hearing aid 1OB may be used not only to exchange the cross- talk compensation signals, but also to exchange audio signals used for acoustic beam forming, noise reduction and/or auditory scene classification, see e.g. V. Hamacher, U. Kornagel, T. Lotter, H.Puder: "Binaural signal processing in hearing aids", in "Advanced in Digital Speech Transmission", R. Martin, U. Heute, C. Antweiler (eds.), p. 401-30, Wiley, 2008.

In Fig. 2 a signal processing model of an example of a bilateral bone conduction hearing aid fitting according to the invention is shown, according to which sound is picked up at the right ear by the microphone 12A of the right ear hearing aid 1OA and at and the left ear by the microphone 12B of the left ear hearing aid 1OB, respectively. For simplification, a time discrete signal processing is assumed, so that the z-transform can be used to represent the signals in the "frequency domain". The audio signals captured by the microphones are then represented by Xi(z) and X ₂ (z), respectively. G ₁ (Z) and G ₂ (z) represent a the transfer function of digital filter for amplification and frequency shaping (these filters correspond to the right ear audio signal processing unit 14A and the left ear audio signal processing unit 14B, respectively). The signals Xi '(z) and X ₂ ' (z) result from applying these filters to X(z) and X ₂ (z), respectively. The transfer functions of the output transducers 16A and 16B are designated by Si (z) and S ₂ (Z), respectively, and the resulting bone vibration signals at the transducer contact points are designated by Yi (z) and Y ₂ (z), respectively.

The cranial transfer functions from the transducer contact points to the ipsilateral cochlea 20A, 20B are represented by Bn(z) and B ₂₂ (Z), respectively, while the transcranial transfer functions from the transducer coupling points to the contralateral cochlea 20B and 20A are designated by Bi ₂ (z) and B ₂ i(z), respectively, with the transcranial transfer functions describing the transfer functions of the cross-talk paths. The sum of the sound arriving from the ipsilateral ("wanted") and the contralateral ("unwanted cross talk") transducer at the particular cochlea 20A or 20B is described by Zi (z) and Z ₂ (z), respectively.

According to S. Stenfelt and R.L. Goode: "Transmission properties of bone conducted sound: Measurements in cadaver heads", JASA 118(4), p. 2373-91, the transmission of vibration in the skull below the first skull resonance frequency, which is approximately I kHz, can be approximated by a linear system; for higher frequencies it is not clear whether the bone conduction by the skull can be modeled by a digital filter. However, it is well-known that the frequencies below 1 kHz significantly contribute to binaural hearing benefits, so that a solution reducing cross talk below 1 kHz would be beneficial. It is the object of the invention to eliminate, as far as possible, the cross- talk signals caused by the transcranial transfer functions Bi ₂ and B ₂₁ . To this end, the right ear hearing aid 1OA is provided with a filter unit 18A providing for a transfer function C ₁ (Z), and the left ear hearing aid 1OB is provide with a filter unit 18B providing for a transfer function C ₂ (z). The filter unit 18A provides for a right ear cross-talk compensation signal, and the filter unit 18B provides for a left ear cross-talk compensation signal, respectively, which signal is combined with the respective contralateral processed audio signal Xi '(z) and X ₂ ' (z). In practice, the cross- talk compensation signals are negative, so that the respective cross talk compensation signal actually is subtracted from the respective contralateral processed audio signal in order to generate the output signal supplied to the transducer 16A and 16B, respectively.

The signals Zi (z) and Z ₂ (z) arriving at the right ear cochlea 2OA and the leftear cochlea 20B, respectively, is given by (in the following "z" will be omitted for simplification):

Z ₁ = B ₁₁ S ₁ [G ₁ X^ C ₂ G ₂ X ₂ ] + B ₂₁ S ₂ [G ₂ X ₂ + C ₁ G ₁ X ₁ ]

= X ₁ G ₁ [B ₁₁ S _{1 +} B ₂₁ S ₂ Q] ₊ X ₂ G ₂ [B ₂₁ S _{2 +} B ₁₁ S ₁ C ₂ ] (1) Z ₂ = B ₂₂ S ₂ [G ₂ X _{2 +} C ₁ G ₁ X ₁ ] + B ₁₂ S ₁ [G ₁ X _{1 +} C ₂ G ₂ X ₂ ]

= X ₂ G ₂ [B ₂₂ S _{2 +} B ₁₂ S ₁ C ₂ ] + X ₁ G ₁ [B ₁₂ S _{1 +} B ₂₂ S ₂ C ₁ ] (2)

If the Filters Ci(z) and C ₂ (z) are chosen to:

Ci = - [Bi ₂ Si] / [B ₂₂ S ₂ ] (3)

C ₂ = - [B ₂₁ S ₂ ] Z [B ₁₁ S ₁ ] (4)

the cross- talk is cancelled out, i.e. both cochlear receive only bone conducted signals coming from the ipsilateral transducer.

For Zi (z) and Z ₂ (Z) one then finds:

Zi = GiSiBu[I - (B _2I Bi ₂ V(B _I iB ₂₂ )] Xi (5)

Z ₂ = G ₂ S ₂ B ₂₂ [I - (B _2I Bi ₂ V(B _I iB ₂₂ )] X ₂ (6)

In other words, for generating the cross-talk compensation signals not only the estimated transcranial transfer function but also the estimated ipsilateral cranial transfer function is taken into account. In particular, the right ear cross-talk compensation signal may be generated by amplifying the processed right ear audio signals, i.e. the output signals of the right ear audio signal processing unit 14A, by a factor corresponding to the ratio of the cranial transfer functions from the right ear output transducer 16A to the right ear cochlea 2OA and the transcranial transfer functions from the left ear output transducer 16B to the right ear cochlea 2OA, multiplied by the ratio of the right ear output transducer transfer function to the left ear output transducer transfer function. The left ear cross-talk compensation signal is generated analogously.

These transfer functions B ₁₁ , B ₁₂ , B ₂₂ and B ₂₁ may be estimated by picking up bone conduction sound reaching the right ear cochlea 2OA and bone conduction sound reaching the left ear cochlea 20B by using vibration sensors, such as accelerometer sensors, 34A and 34B attached to the skull on the mastoid at a position as close to the respective cochlea 20A, 20B as possible, and wherein the bone conduction sound is generated by the right ear output transducer 16A and the left ear output transducer 16B, respectively. Since the transfer functions B ₁₁ , B ₁₂ , B ₂₂ and B ₂₁ usually do not change, the accelerometer sensors, 34A and 34B are removed after the fitting procedure.

The best measurement position, of course, would be the respective cochlea 20A, 20B itself. However, in view of the relatively large wavelength of bone conducted sound, the cross-talk cancellation effect provided by the present invention at a place quite close to the cochlea should not deviate too much from the effect at the cochlea itself.

For the calculation of the transfer functions, which has to deal with stability, causality and delay issues, signal processing techniques known from virtual audio imaging can be applied, such as techniques described in J. Kim, S. Kim, C. Yoo, "A Novel Adaptive Crosstalk Cancellation using Psychoacoustic Model for 3D Audio", Proceedings Acoustics, Speech and Signal Processing, 2007, ICASSP 2007, Vol. 1, p. 1185-1- 188. The transcranial transfer function B ₁₂ , B ₂₁ and the cranial transfer function B ₁₁ , B ₂₂ for each of the ears may be estimated by using the both output transducers 16A, 16B, the ipsilateral vibration sensor (which is in case of the left ear the sensor 34B), and the contralateral processed audio signals (in this case the audio signals generated by the right ear audio signal processing unit 14A from the audio signals captured by the right ear microphone arrangement 12A), while the ipsilateral audio signal processing unit (here the left ear unit 14B) is not involved. Then the cross- talk compensation signal provided by the contralateral filter unit (here the right ear unit 18A) is adjusted so as to minimize the signal picked up by the ipsilateral vibration sensor 34B. For minimizing the signal picked up by the ipsilateral vibration sensor 34B a least mean squares (LMS) algorithm may be used. An example of such measurement configuration is shown in Fig. 3 for the left ear; the set-up of Fig. 3 involves the transfer functions B ₁₂ and B ₂₂ for determining the transfer function C ₁ of the right ear filter unit 18A. The desired transfer function C ₂ of the filter unit 18B of the left ear hearing aid may be determined by an analogous set-up using the the right ear vibration sensor 34A.

Alternatively, standard filters based on empiric cranial transfer function data averaged across a large group of persons, , i.e. "default filters" based on measured transfer functions averaged across a large group of persons, may be used for determining the transfer functions of the filter units 18 A, 18B.

In addition, after minimizing the signal picked up by the ipsilateral vibration sensor as described with regard to Fig. 3, the transfer functions C ₁ , C ₂ of the filter units 18A and 18B, i.e. the cross-talk compensation signals, may be further adjusted by loudness measurements, so as to minimize loudness perception in the ipsilateral ear.

Alternatively, the cross- talk compensation signal may be adjusted so as to minimize the measured vibrations of the middle ear ossicles, of the oval window or of the round window. Such vibration measurements may be performed in a non-invasive manner by using, for example, a Laser-Doppler-vibrometer through the tympanic membrane.

It is also to be noted that the filter units 18A and 18B attenuate the ipsilateral signals by the factors [1 - (B ₂₁ B ₁₂ V(B ₁₁ B ₂₂ )]. Since |B ₁₂ | < IB _{1 1} I and |B ₂₁ | < IB ₂₂ I and since there are phases differences in B ₂ vs. Bi ₁ and B _2I vs. B ₂₂ , the attenuation factor can be small but never be zero, so that it can be compensated by applying the additional gain [1 - (B _2I Bi ₂ )Z(B _I iB ₂₂ )] ^"1 to Gi (z) and G ₂ (Z) respectively. Preferably, such attenuation is compensated by applying an appropriate additional gain in the audio signal processing unit 14A and 14B.

The measurement set-up of Fig. 3 may be realized by transfer function estimation units 36A and 36B provided in the right ear hearing aid 1OA and the left ear hearing aid 1OB, respectively. The right ear transfer function estimation unit 36A receives the signals from the contralateral vibration sensor 36B and generates a corresponding signal for adjusting the ipsilateral filter unit 16A. Accordingly, the left ear tansfer function estimation unit 36B receives the signals from the contralateral vibration sensor 34A and generates a corresponding signal for adjusting the ipsilateral filter unit 18B. The system also generates control signals for turning off the respective contra- lateral audio signal processing unit 14A, 14B during transfer function measurements.

In general, the above -described principle of cross-talk compensation may be applied also to bilateral system comprising a bone conduction transducer only on one side / ear, while at the other side / ear a type of output transducer other than bone conduction is used, such as a loudspeaker.

In Fig. 4 a modification of the system of Fig. 1 is shown, wherein the left ear bone conduction output transducer 16B is replaced by a left ear output transducer 116B formed by an electro- acoustic transducer (loudspeaker) or an electro- mechanical output transducer which is mechanically directly coupled to the eardrum, the ossicular chain or the cochlea 2OB of the left ear, such as an active middle ear implant or a DACS (direct acoustic cochlea stimulation) device (of course, the role of the right ear and the left ear could be interchanged).

Such type of output transducer 116B does not provide for a significant cross- talk signal to the other (right) cochlea 2OA (i.e. the transcranial transfer function B ₂₁ of Fig. 1 is very small). Therefore it is not necessary to provide for a cross-talk compensation signal from the (left ear) hearing aid 1OB to the other (right ear) hearing aid 1OA so that the (left ear) hearing aid 1OB does not need to have the filter unit 18B which is used in the example of Fig. 1 for generating a left ear cross- talk compensation signal (and the elements 34A, 36B used in Fig.l for estimating the transcranial transfer function B ₂₁ ).

The impact of the cross-talk compensation signal on the gain of the left ear hearing aid 1OB has to be compensated in the manner discussed above with regard to the system of Fig. 1.