Filter bank system for hearing aids

p0001The present invention relates to a filter bank system for a hearing aid with an analysis filter bank for splitting an input signal into sub-band signals, processing means for amplifying at least one of the sub-band signals and a synthesis filter bank for synthesizing the processed sub-band signals into an output signal.

p0002Hearing devices are portable hearing devices used to support the hard of hearing. To meet the numerous individual requirements, different types of hearing devices such as behind-the-ear (BTE) hearing aid with an external receiver (RIC: receiver in the canal) are and In-the-ear (ITE) hearing, as well as concha hearing aids or canal hearing aids (ITE, CIC). The hearing by way of example are worn on the outer ear or in the ear canal. In addition, be on the market are bone, implantable or vibrotactile hearing aids available. The stimulation of the damaged hearing is either mechanically or electrically.

p0003Hearing devices principally as essential components an input transducer, an amplifier and an output transducer. The input transducer is generally a sound receiver, eg. As a microphone, and / or an electromagnetic receiver, eg. As an inductor. The output transducer is usually implemented as an electroacoustic transducer, eg. As a miniature loudspeaker, or as an electromechanical transducer, eg., Bone conduction hearing aid. The amplifier is usually integrated into a signal processing unit. This basic structure is in<figref idrefs="f0001">1 shows</figref> illustrated by the example of a behind-the-ear hearing aid. In a hearing aid housing 1 to be worn behind the ear, one or more microphones 2 for recording the ambient sound are built. A signal processing unit 3, which is also integrated in the hearing aid housing 1, processes the microphone signals and amplifies them. The output of the signal processing unit 3 is transmitted to a loudspeaker or receiver 4, which outputs an acoustic signal. The sound is optionally transmitted through a sound tube which is fixed with an earpiece in the ear canal to the eardrum of the wearer. The power supply of the hearing device and especially to the signal processing unit 3 by a likewise integrated into the hearing aid housing 1 battery. 5

p0004Sound signals, which are received by one or more microphones of a hearing aid are usually disassembled for further processing in sub-band signals. For this purpose, usually one uses one or more frequency-selective digital analysis filter banks (AFB), whereby K> 1 sub-band signals are obtained. After the decomposition of sub-band specific signal manipulations can be performed. With hearing aids, these are particular gains in the individual sub-bands. A subsequent re-synthesis of the manipulated sub-band signals can be carried out by means of a digital synthesis filter bank (SFB).

p0005On quality filter banks in hearing certain requirements are imposed. Thus, a channel bandwidth of about 250 Hz is required for example in the lowermost bands. Otherwise the bandgap should align roughly to the Bark scale. Furthermore, a number of channels of at least 22 is desirable. Spurious aliasing should certainly be below 60 dB. Due to the intensive sub-band processing (especially the high gain required to compensate for the hearing loss) hearing aids are conventional methods for eliminating aliasing not effective. The filter banks are therefore in principle "not critical" scan. Furthermore, should the group delay (both for AFB and SFB) are well below 5 ms and the group delay distortions do not exceed a certain extent. Especially for high frequencies is the group delay to be kept as low as possible, which is a major limiting factor for the filter bank.

p0006The publication <patcit id="pcit0001" dnum="DE69833749T2"><text>DE 698 33 749 T2</text></patcit> discloses a filter bank system and method for filtering and separating an information signal in different frequency bands for audio signals in hearing aids. In an analysis filter bank, an input signal is divided into subband signals. The subband signals are optionally amplified and assembled in a synthesis filter bank to an output signal. In the analysis filter bank, the signals are oversampled.

p0007The object of the present invention is to make the sub-band processing in hearing aids less computationally intensive.

p0008This object is achieved by a filter bank system for a hearing aid with an analysis filter bank for splitting an input signal into sub-band signals, processing means for amplifying at least one of the sub-band signals and a synthesis filter bank for synthesizing the processed sub-band signals into an output signal, wherein the sub-band signals in the analysis filter bank, the processing device and the synthesis filter bank is oversampled with an oversampling factor U and with respect to the input signal downsampled by a downsampling factor, the stop band attenuation of the individual filters of the analysis filter bank are at least as high as a predetermined first signal-to-noise ratio is increased by a first attenuation value and a function of the oversampling factor U of represents downsampling, and / or the synthesis filter bank are at least as high as a predetermined second signal-noise ratio increased by a second attenuation value which is a function of the oversampling factor U, the downsampling factor and the gain of the processing device, and the magnitude frequency response of the synthesis filter bank is approximately the magnitude frequency response the analysis filter bank is designed in such a way that the complex-valued frequency response of the cascade of analysis filter bank and synthesis filter bank, bypassing the processing device is approximately an all-pass transfer function.

p0009Advantageously, it is possible by the specific stop band attenuation of the individual filters and the amount frequency adjustment to reconstruct the downsampled subbands in spite of hearing aid typical gains at a relatively distortion-free output.

p0010Preferably, the AFB and SFB are each implemented as a multistage filter banks for the gradual increase in the downsampling or Aufwärtstastfaktors. This can be in the individual subbands different Abwärtstastfaktoren use.

p0011Specifically, the AFB and / or SFB can be implemented as a polyphase filter banks, wherein all the individual frequency characteristics of the filter banks are each derived from a prototype filter by modulation. Thus, the filter banks can easily design.

p0012In a further development, the prototype filter of the AFB compared to the prototype filter of the SFB, a different stop band attenuation and / or a different damping and / or a different attenuation in the passband own. This is not the same high quality blocking attenuation must be carried out in two filter banks.

p0013Furthermore, the AFB and the SFB can each represent non-uniform filter banks. This can be in the sub-band signals different sampling or oversampling factors realized.

p0014At least one of the filter banks or a part-filter bank of the filter banks may be minimum phase or linear. In the case of minimum-each filter bank has a small group delay and in the case of linear-constant group delay.

p0015Likewise, the prototype filter used for the filter banks, to be minimum phase or linear. This also leads to a small or constant group delay.

p0016Furthermore, substantially constant group delay of the individual filters of the filter banks in each passband. In addition, the group delay of the filters of the filter banks in a substantial part of the transition region from the passband can be constant for the stop band. The larger the area of ​​the constant group delay, the less distortion can be expected.

p0017According to a particularly preferred embodiment, each U polyphase branch filters (U = oversampling) of the prototype filter are supplied with the same signal. This can be in the implementation of the filter banks numerous latches save.

p0018The present invention is explained in more detail with reference to the accompanying drawings, in which:<dl id="dl0001"><dt>1 shows</dt><dd>the basic design of a hearing device according to the prior art;</dd><dt>FIG 2</dt><dd>a filter bank system according to the present invention; </dd><dt>FIG 3</dt><dd>the magnitude frequency response of the bandpass filter of the filter banks;</dd><dt>FIG 4</dt><dd>the magnitude frequency response of the filter bank system and</dd><dt>FIG 5</dt><dd>the construction of a filter with reduced memory number.</dd></dl>

p0019The more detail below embodiments represent preferred embodiments of the present invention.

p0020In <figref idrefs="f0002">FIG 2</figref> is a filter bank system, here a filter bank cascade consisting of a multi-stage analysis filter bank (AFB) and a multi-stage synthesis filter bank (SFB) shown. The exemplary filter bank cascade used for signal processing in a hearing device and in particular in a hearing aid. The input-side filter bank (FB1) of the AFB decomposes the input signal into four channels. The output side Fi lterbänke FB2A, FB2B, FB2C and FB2D disassemble the four channels further into ultimately 24 channels. The bottom channel of the FB1 is thereby broken by the FB2A in twelve channels, while the other three channels of FB1 be decomposed using the output-side filter banks FB2B, FB2C and FB2D each in four channels. The input sample rate of FB1 is for example 24 kHz. The sampling rate between the two filter bank stages f_ZW is 6 kHz in the example chosen. The sampling in the sub-band channels at the output of the AFB is in the high frequency groups so after the filter banks FB2B, FB2C and FB2D each 3kHz. The sample after the filter bank FB2A the lower frequency group is 1.2 kHz. It takes place here advantageously a downsampling by downsampling 8 and 20 respectively.

p0021After AFB a sub-band-specific signal manipulation is carried out in particular gain which in <figref idrefs="f0002">FIG 2</figref> but is not shown. For clarity, joins the AFB in<figref idrefs="f0002">FIG 2</figref> immediately the SFB for re-synthesis of the signal. The CRC is constructed with respect to the filter banks in the individual stages symmetrical with the AFB. Thus are in the lowest stage of the SFB filter banks FB3A, FB3B, FB3C and FB3D which together twelve or four sub-band signals to a signal. The four resulting signals with a sampling rate of 6 kHz are the higher stage of the synthesis FB4 supplied composing the signals into an output signal with a sampling rate of 24 kHz.

p0022The wider filter banks FB2A and FB3A in the lower frequency group also lead here τ to increased group delay_G opposite the next higher frequency group with the narrower filter banks FB2B and FB3B. Therefore, the filter bank is followed FB3B an equalization filter (equalizer EQ). This equalization filter EQ increases the group delay of the filter bank FB3B on the upper (high frequency) band edge to the value of the group delay of the filter bank FB3A at its lower band edge.

p0023It is further <figref idrefs="f0002">FIG 2</figref> shows that a Übertastfaktor U = 2 is selected in the first stage filter bank FB1 the analysis filter bank. The Übertastfaktoren in the second stage be 4 or 5. With the filter bank structure so chosen is an approximately 12 kHz wide input signal into 12 channels of bandwidth 240 Hz (channels 1 to 12) and in 12 channels of bandwidth can, for example, 750 Hz (channels 13 to 24) decompose.

p0024In the <figref idrefs="f0002">FIG 2</figref> illustrated filter bank structure of the analysis filterbank is optimized with regard to group delay and power consumption. This AFB filter bank structure has an input side and an output-side filter bank based on a given filter type and can be determined as follows. The input-side filter bank has a variable first parameter, a channel number and a variable second parameter an oversampling. For the optimization of the multistage filter bank now has a group delay and an operation rate is determined for each of a plurality of pairs of values ​​of the first and second parameters. From the value pairs the one is selected, in which the associated group delay and the associated operation rate satisfy a predetermined criterion, in particular as low as possible. The input-side filter bank is then configured with the channel number and the oversampling factor corresponding to the selected pair of values. The SFB is constructed symmetrically to the AFB, with the exception of optionally employable equalization filter EQ.

p0025The subband signals in channels 1 to 24 of the AFB, the SFB and the intervening processing means are therefore scanned with a predetermined oversampling factor U and with respect to the input signal by a downsampling factor (here 20 in the channels 1 to 12 and 8 in the channels 13 to 24 ) down sampled.

p0026The individual filters of the AFB have a sufficiently high stop band attenuation in order to prevent nonlinear signal distortions by aliasing. In particular, the stop-band attenuation is at least as high as a desired signal-to-noise ratio. To this end, in<figref idrefs="f0003">FIG 3</figref> recorded the magnitude frequency response of a filter. For example, if the signal-to-noise ratio should be 60 dB and no other problems, are given by aliasing, the stop band attenuation of the filter may be 60 dB. but are expected to further distortions by aliasing components, so the blocking attenuation must be increased accordingly. For example, the stop band attenuation can be increased per expected aliasing component by 3 dB. The expected number of aliasing components can be determined from the oversampling and the downsampling in the usual way. In the example of<figref idrefs="f0003">FIG 3</figref> are expected three additional aliasing components, so that the stop band attenuation of 3 x 3 = 9 dB dB is increased to a total of 69 dB.

p0027Additionally or optionally available as an alternative, the filter of the SFB must be equipped with sufficiently high damping. Here again the starting point of the desired signal-to-noise ratio, but it can be different from the signal-to-noise ratio of the filters of the AFB. The stop band attenuation is again increased by a suitable damping value that results from the oversampling and the downsampling. In the SFB should however be borne in mind that in the hearing aid signal processing means at least some sub-bands signal amplification is provided. Therefore, undesirable components, especially imaging components at the upsampling result, will be strengthened accordingly. This also needs to be sufficiently damp. Consequently, the stop-band attenuation of the filter in the SFB is also increased depending on the gain of the processing device. Thus, non-linear signal interference by imaging the SFB can be avoided. Phase matching (phase matching) the SFB to the AFB for the elimination of aliasing will however does not take place. This is when a hearing aid application does not make sense.

p0028In addition, the magnitude frequency response of the SFB is approximately matched to the magnitude frequency response of the AFB in the present invention. Thus, the complex-valued frequency response of the cascade of AFB and SFB obtained bypassing the processing means approximately an all-pass transfer function. The word "approximately" refers to the usual concept of NPMR (Near Perfect Magnitude Reconstructions). It is here so no PMR (Perfect Magnitude Reconstructions), since this may not be due to the gain in subbands of the hearing aid. The PMR would namely therefore lead to a transparent system with the constant transmission factor 1 without amplification. Thus, it can not otherwise customary for filter banks with perfect reconstruction corresponding aliasing compensation, which can be derived purely mathematically, be chosen. Rather, an appropriate oversampling and a sufficiently high isolation should be selected to avoid interference. The magnitude frequency response of the filter bank system is it to optimize based accordingly.

p0029<figref idrefs="f0003">FIG 4</figref> shows the magnitude frequency response |<i>H_FBS</i>| a filter bank system, in which the magnitude frequency response of the SFB with the magnitude frequency response of the AFB is coordinated, so that there is approximately an all-pass when the hearing aid gain 1. Concretely shows<figref idrefs="f0003">FIG 4</figref> the magnitude frequency response of a hearing aid to Hochtonverstärkung. found only in the higher channels here instead of a gain. If this gain cut off, a transparent system with approximately the same gain 1 dashed curve results over the entire frequency range (see. In the higher frequency range<figref idrefs="f0003">FIG 4</figref>). Through the filter bank structure, however, give some variation of the amount of frequency response. There is therefore no mathematically perfect reconstruction of the input signal. Rather, the filter bank cascade reconstructs the magnitude spectrum of the input signal only approximately perfectly. The range of variation is for example 2 dB.

p0030Below variants of the filter bank system according to the invention are presented, which can lead to further improvements. Thus, in principle, the AFB and / or the SFB as a single- or multi-stage filter bank to be realized for the gradual reduction or increase of the sampling rate. It is also advantageous if the AFB and / or the SFB are realized as (complex or real modulated) polyphase filter banks, where all the individual frequency responses of the filter banks are each derived from a prototype filter by modulation. For expenses cheap realization of filter banks the computationally cheap uniform, complex modulated (discrete Fourier transform) polyphase filter banks (see. Heinz Göckler and Alexandra Groth "multirate systems" Schlemmbachverlag, Willburg Sites, 2004) are used with oversampling by the factor U. These polyphase filter banks only a prototype filter for the filter bank channels is needed, and the frequency shift is implemented by means of efficient FFT implementation of the DFT.

p0031Furthermore, the AFB and SFB can be implemented with the same or different bandwidth of the individual filter bank channels as a uniform or non-uniform filter banks, and in the case of non-uniform filter banks, the subband signals may have different sampling rates and / or different oversampling factors. In addition, the prototype filter used for AFB, another stop-band attenuation and / or a different frequency response in the passband have the prototype filter of the SFB.

p0032At least one of the (sub) filter banks the filter bank system can be minimum, ie the zeros of the Z-transform of the filter bank transfer function are located inside the unit circle. This has the advantage that the group delay is relatively small then. Also the prototype filter itself can be minimum. It is also favorable when the group delay of the prototype filter or the corresponding filter bank in the filter passband and also in parts of the filter transition region is approximately constant, if appropriate, to the stop band. As a result, any distortions are reduced.

p0033At least one of the (sub) filter banks or a prototype filter of the filter banks may alternatively be linear phase. This results in a constant group delay, which may be may be beneficial.

p0034Further, the filter banks and the prototype filter non-recursive (FIR) filters or recursive (IIR) filter can be realized. Optional can be the filter banks or prototype filter also partly from the FIR and partly IIR type. In the case of IIR filters, the frequency response of the minimum phase Butterworth filter or a minimum-phase inverse Chebychev filter of type 2 may correspond.

p0035Particular advantages result in the use of oversampling by the factor U of saving latches in the realization of the filter banks with FIR prototype filters. Namely, each U polyphase branch filters h₁, H₅, H₉, ..., H₄₅ of the FIR prototype filter with the same input signal (such as one of the polyphase pk1, pk2, pk3 and PK4 of the input signal) supplied (see FIG. <figref idrefs="f0004">FIG 5</figref>). For each U branch filters to the same set of latches (in<figref idrefs="f0004">FIG 5</figref> each with z₀^-1 in) recourse. Thus, instead of U signal memory chains for U branch filters only a single store chain is required.

p0036In the specific case of the example of <figref idrefs="f0004">FIG 5</figref> is the oversampling factor of two because of the four polyphase pk1 to pk4 eight components are recovered according to the inputs 0 to 7 of the IDFT block for the inverse discrete Fourier transform. Specifically, as in<figref idrefs="f0004">FIG 5</figref> is shown, from the polyphase component pk2 with a single latch chain (eg₀^-1Chain) won the first and fifth component of the IDFT. The same reduction is possible with the appropriate filters the SFB as their structure in principle by transposition (see. Heinz Göckler and Andrea Groth, supra) is traceable to AFB structures.