A method for low-delay signal processing in a hearing instrument

Abstract

The invention discloses a method for low-delay signal processing in a hearing in- strument (1), wherein a first input signal (E1) is generated from an environment sound (4) by means of a first input transducer (M1) of the hearing instrument (1), wherein a digital first intermediate signal (x1) is generated from the first input sig- nal (E1), said first intermediate signal (x1) having a first sample rate (fS,1), wherein in an analysis path (20), a decimation (24) is applied to the first intermediate signal (x1), resulting in a first decimated signal (d1) having a second sample rate (fS,2) smaller than the first sample rate (fS,1), and by means of a signal analysis of the first decimated signal (d1), first filter coefficients (c1j) of a signal processing filter (16) for the first intermediate signal (x1) are calculated, wherein in a processing path (10), the signal processing filter using (16) the first filter coefficients (c1j) is applied to the first intermediate signal (x1) at the first sample rate (fS,1), generating a first processed signal (y1), and wherein a first output signal (A1) of the hearing instrument (1) is derived from the first processed signal (y1).

Description

[0001] Description

[0002] A method for low-delay signal processing in a hearing instrument

[0003] The invention is related to a method for low-delay signal processing in a hearing instrument, wherein a digital intermediate signal is generated from an input signal generated by an input transducer of the hearing instrument, said intermediate signal having a determined sample rate.

[0004] Generally, a hearing instrument according to the invention is understood as meaning any device or system which provides an output signal that can be perceived as an acoustic signal by a user, or which contributes to providing such an output signal, and which, in particular, may have means customized to compensate for an individual hearing loss of the user or to contribute to said compensating for the hearing loss of the user. These are considered as “hearing aids in the strict sense”. However, some devices whose main aim is not to compensate for a hearing loss, may also be regarded as hearing instruments, for example consumer electronic devices (televisions, hi-fi systems, mobile phones, MP3 players etc.), and which preferably may have measures for supporting an individual user’s hearing even in case of no hearing loss. Hearing instruments may be worn on the body or by the ear, and in particular on or in the ear, and may be fully or partially implanted.

[0005] Within the present context a hearing aid “in the strict sense” can be understood as a small, battery-powered, microelectronic device designed to be worn behind or in the human ear by a hearing-impaired user. Prior to use, the hearing aid is adjusted by a hearing aid fitter according to a prescription. The prescription is based on a hearing test, resulting in a so-called audiogram, of the performance of the hearing- impaired user’s unaided hearing. The prescription is developed to reach a setting where the hearing aid will alleviate a hearing loss by amplifying sound at frequencies in those parts of the audible frequency range where the user suffers a hearing deficit. A hearing instrument comprises one or more microphones, a battery, a microelectronic circuit comprising a signal processor, and an acoustic output transducer. The signal processor is preferably a digital signal processor. The hearing instrument is enclosed in a casing suitable for fitting behind or in a human ear.

[0006] Hearing loss of a hearing impaired person is quite often frequency-dependent. This means that the hearing loss of the person varies depending on the frequency. Therefore, when compensating for hearing losses, it can be advantageous to utilize frequency-dependent amplification. Hearing aids therefore often provide to split an input sound signal received by an input transducer of the hearing aid, into various frequency intervals, also called frequency bands, which are independently processed. In this way, it is possible to adjust the input sound signal of each frequency band individually to account for the hearing loss in respective frequency bands. The frequency dependent adjustment is normally done by implementing a band split filter and compressors for each of the frequency bands, so-called band split compressors, which may be summarised to a multi-band compressor. In this way, it is possible to adjust the gain individually in each frequency band depending on the hearing loss as well as the input level of the input sound signal in a specific frequency range. For example, a band split compressor may provide a higher gain for a soft sound than for a loud sound in its frequency band.

[0007] The filter banks used in such multi-band compressors are well known within the art of hearing aids, but are nevertheless based on a number of tradeoffs. Most of these tradeoffs deal with the frequency resolution as will be further described below.

[0008] There are some very clear advantages of having a high resolution filter bank. The higher the frequency resolution, the better individual periodic components can be distinguished from each other. This gives a much finer signal analysis and enables more advanced signal processing. Especially noise reduction and speech enhancement schemes may benefit from a higher frequency resolution. However, a filter bank with a high frequency resolution generally introduces a correspondingly long delay, which for most people will have a detrimental effect on the perceived sound quality. This also holds for hearing instruments (i.e. , “in the broader sense”) whose main function is not the compensation of a hearing loss, but which do provide certain hearing support for a user, e.g., in complex hearing situations in order to increase the signal-to-noise-ration of a useful signal and / or to improve an intelligibility of a useful signal.

[0009] It is therefore an object of the present invention to provide a method of operating a hearing instrument that provides improved low delay signal processing.

[0010] SUMMARY OF THE INVENTION

[0011] According to the invention, this object is solved by a method for low-delay signal processing in a hearing instrument, wherein a first input signal is generated from an environment sound by means of a first input transducer of the hearing instrument, wherein a digital first intermediate signal is generated from the first input signal, said first intermediate signal having a first sample rate, wherein in an analysis path, a decimation is applied to the first intermediate signal, resulting in a first decimated signal having a second sample rate smaller than the first sample rate, and by means of a signal analysis of the first decimated signal, first filter coefficients of a signal processing filter for the first intermediate signal are calculated, wherein in a processing path, the signal processing filter using the first filter coefficients is applied to the first intermediate signal at the first sample rate, generating a first processed signal, and wherein a first output signal of the hearing instrument is derived from the first processed signal. Embodiments of particular advantage, which may be inventive in their own right, are outlined in the depending claims and in the following description.

[0012] The notion of a hearing instrument generally comprises any device which is configured to generate a sound signal from an electrical signal and to present it to the hearing of a wearer of this device, i.e. in particular headphones (e.g. as an "earbud"), a headset, data glasses with loudspeakers, etc. However, a hearing instrument also includes a hearing aid in the strict sense, i.e. a device for providing support for a hearing impairment of a user (i.e., the hearing aid wearer), wherein an input signal generated from an environment sound by means of a microphone is processed into an output signal and in particular amplified frequency band-wise, and an output sound signal generated from the output signal by means of a loudspeaker or the like is suitable for at least partially compensating for the hearing impairment of the wearer, in particular in a user-specific manner.

[0013] In the particular case of a hearing aid as the hearing instrument, any mechanical design such as a Behind-The-Ear (BTE) device, a Receiver-ln-The-Ear (RITE) device, a Receiver-In-Canal (RIC) device, an In-The-Ear (ITE) device, or a Com- pletely-ln-Canal (CIC) may be comprised in notion of the hearing instrument.

[0014] The notion of an input transducer, such as the first input transducer, in particular may comprise any device that is configured to generate a corresponding electrical signal from a sound signal. In particular, when the input signal is generated by the corresponding input transducer, pre-processing can also take place, e.g. in the form of a linear pre-amplification. The input signal generated this way, in particular, may be given by an electrical signal whose current and / or voltage fluctuations essentially represent the sound pressure fluctuations of the air in the environment that yields the environment sound.

[0015] In the present context, the generation of a resulting signal from an (electrical) source signal shall in particular comprise that the resulting signal “generated from” the source signal only contains signal contributions (e.g., acoustic samples or frames of samples) from said source signal, and any other signal than said source signal is, if at all, only used for controlling the generation of said resulting signal, i.e., the signal processing (e.g., via a control of a compression or amplification, possibly frequency band-wise, that is applied to the source signal in order to generate the resulting signal, said control being possibly dependent on said other signal). Likewise, the derivation of a resulting signal from an (electrical) source signal in the present context shall in particular comprise that the resulting signal “derived from” the source signal contains signal contributions from said source signal, and may also contain signal contributions from another signal, e.g., via directional processing of the source signal and said other signal.

[0016] The first intermediate signal may be generated from the first input signal directly by an analog-to-digital (A / D) conversion of the first input signal at the first sample rate for the first input signal being an analog signal. However, for the first input signal already being a digital signal, the first intermediate signal may be generated from the first input signal by digital signal processing which may comprise a preliminary decimation (in this case, an initial sample rate of the first input signal is greater than the first sample rate of the first intermediate signal) and / or by a beamforming process with another input signal. Furthermore, in the latter case, the generation of the first intermediate signal from the first input signal in particular may be given by a trivial identity signal processing, i.e. , the first input signal is directly taken as the first intermediate signal.

[0017] The signal processing of the first intermediate signal is then split into a processing path, in which the actual processing of the intermediate signal, i.e., the modification of its signal contributions, takes place, and an analysis path, in which filter coefficients for said modification in the processing path are calculated on the basis of the intermediate signal itself.

[0018] In the analysis path, a decimation is applied to the respective copy of the intermediate signal in said path, i.e., the time resolution of the first intermediate signal in the analysis path is reduced, preferably, by an integer factor, and the first sample rate of the first intermediate signal is reduced accordingly to the second sample rate of the first decimated signal resulting from said decimation.

[0019] Preferably, a decimation in the present context in general includes a downsampling step of the corresponding signal to be decimated, as well as a low-pass filtering step to remove undesired high-frequency content beyond the new Nyquist frequency of the downsampled signal. This low-pass filtering process may help to reduce aliasing artifacts that can occur during the downsampling process. By performing said decimation, it can be achieved that, under reasonable assumptions, only the desired frequency range (i.e. , below the decimated sample rate or twice the Nyquist frequency resulting from said decimation) is retained, reducing thus the possibility of distortions or artifacts. Work directly with the reduced sampling frequency from downsampling without a “proper” decimation including the low- pass filtering mentioned above might give rise to risk introducing aliasing artifact and a potential corruption of the signal.

[0020] The decimation of a signal generally reduces the sample rate, which in turn reduces the computational load for subsequent signal processing stages. By working with a lower sample rate, computationally efficient algorithms designed for the reduced rate may be applied. This can lead to significant computational savings compared to processing the signal at the original higher sample rate. Furthermore, decimation possibly allows for adjusting the decimation factor (i.e., the ratio of the original sample rate to the decimated sample rate) based on the specific requirements of the following signal processing application(s). This flexibility allows for striking a balance between computational efficiency and preserving important signal information of the signal to be decimated (and processed afterwards).

[0021] As the minimum filter length of an acoustic processing FIR filter, i.e., the minimum number of samples or minimum frame length needed to be processed in said filter for every resulting sample (i.e., in this case, of the first processed signal), is often given by physical considerations (e.g., the possible periodicity length of the acoustic signal), the signal processing at a given sample rate cannot simply be truncated to lower-length filters without risking to destroy important acoustic information. Thus, due to the convolutional characteristic of FIR filters (which are most commonly used for signal processing in hearing instruments as the ones described above), by using decimation of a signal prior to its processing by a factor of N, a reduction of the complexity of 0(N2) may possibly be achieved.

[0022] In the present context, the first filter coefficients for processing the first intermediate signal at the first sample rate in the processing path, are calculated based on an analysis of the first decimated signal at the second sample rate in the analysis path. So, at first, the signal processing of the first intermediate signal is still performed at the higher first sample rate, and only the analysis is performed at the lower second sample rate. However, even in this case, the computational complexity may be reduced by reducing the number of calculation steps in the analysis path, according to the reduced second sample rate of the first decimated signal used therein.

[0023] In order to obtain the first filter coefficients for the signal processing filter to be applied to the first intermediate signal, corresponding initial filter coefficients are calculated in the analysis path, the initial filter coefficients being the filter coefficients of a signal processing filter for the first decimated signal, with the desired processing properties. Then, the first filter coefficients for the signal processing filter in the processing path (operating at the first sample rate) are preferably obtained from the initial filter coefficients for the signal processing filter in the analysis path (operating at the second sample rate) by an appropriate mapping of the initial filter coefficients to (and in particular onto) the first filter coefficients.

[0024] The signal processing filter using said first filter coefficients, obtained on the basis of the analysis of the first decimated signal in the analysis path, is then applied to the first intermediate signal in the processing path, thereby generating the first processing signal. In particular, the first processed signal only contains signal contributions of the first intermediate signal processed by the signal processing filter (using said first filter coefficients).

[0025] Finally, the first output signal is then derived from the first processed signal, in particular by including all signal contributions of the first processed signal, and possibly signal contributions from another signal (e.g., for directional processing based on the first processed signal and said other signal) in the first output signal.

[0026] It is then essentially a matter of mere definition whether the first output signal is given by an analog signal directly to be converted into an output sound signal by means of an output transducer of the hearing instrument, or whether the first output signal is given by a digital signal, to be directly fed to a digital-to-analog (D / A) converter that generates the analog signal for said output transducer. For the signal processing in hearing instruments, there is always a tradeoff between signal quality, real-time perception and energy efficiency. To ensure a high signal quality, a high sample rate is of advantage, increasing the computational complexity, and thus, the energy consumption due to the increase in calculation steps required per time unit of an output signal. On the other hand, decimation typically introduces a signal delay which in turn may compromise the real-time perception, since the typical filters used in hearing instruments for decimation, such as the low-pass filtering mentioned above, have a non-negligible intrinsic signal delay. Depending on the specific decimation filter, this delay is at least of the order of the inverse bandwidth (i.e., the inverse sample rate) of the decimated signal, and may even be larger since the process may be temporally more complex than just “averaging” N source samples into one resulting sample.

[0027] For a low-delay signal processing with low computational complexity, it is thus suggested to keep the first intermediate signal at its original first sampling rate, thus avoiding an extra delay that would be introduced in the processing path by a decimation, and to apply the decimation only in the analysis path, thereby achieving a reduction of the computational complexity in said analysis path. The tradeoff of a small delay introduced in the analysis path does not noticeably compromise the real-time perception of the first processed signal, and in consequence, of the first output signal, as this delay only applies to the first filter coefficients, but not to the first processed signal itself.

[0028] By an adequate choice of the first filter coefficients, e.g., by an adequate mapping of the corresponding initial filter coefficients corresponding to a signal processing filter for the first decimated signal in the analysis path onto the first filter coefficients of the signal processing filter for the first intermediate signal in the processing path, the computational complexity (reduced due to the decimation in the analysis path) may be reduced even further.

[0029] In summary, using decimation offers the advantage of preserving signal information, reducing computational efficiency, and providing flexibility in adapting the downsampling ratio. These benefits make decimation a preferred approach in audio signal processing to maintain signal integrity and optimize computational resources.

[0030] In an embodiment, in the analysis path, initial filter coefficients of the signal processing filter are calculated by means of the first decimated signal, wherein the first filter coefficients are derived from the initial filter coefficients using a delay in time-domain and / or a number of copies of a transfer function of a filter based on the initial filter coefficients, said number being dependent on a ratio between the first sample rate and the second sample rate, in frequency domain. This comprises in particular: Let the signal processing filter be a finite impulse response (FIR) filter, x1 [n] denote the first intermediate signal in (discrete) time domain, and c1 j the first filter coefficients, such that the application of the signal processing filter with said first filter coefficients to the first intermediate signal x1 [n] yields the first processed signal y1 [n]:

[0031] (i) with the filter length K N, then the first filter coefficients c1 j are derived from initial filter coefficients cinmby a delay, i.e. , by means of delaying the m-th initial filter coefficient cinmby an amount of preferably by m ■ (N - 1 ) time units (or m ■ N time units) or by a similar relation involving the factor m ■ N, being N the decimation ratio defined by the first sample rate fs,i (prior to decimation) divided by the second sample rate fs,2 (after decimation).

[0032] In particular, the first filter coefficients other than the ones obtained by said delay (in time domain) or said number of copies of said transfer function (in frequency domain) from the initial filter coefficients, may be set to zero, i.e.,

[0033] 1> r cinmfor j = m • N, J t 0 else.

[0034] (ii) The reason for this is that in frequency domain, the action of said delay operation on the transfer function of the filter represented by the initial coefficients, is given by generating N copies of said transfer function over the sample window.

[0035] Preferably, said delay corresponds to the inverse of said second sample rate fs,2 (i.e., 1 / fs,2 = N / fs,i = N times the length of the sample of the intermediate signal), and / or said number of copies of said transfer function corresponds to an application of the signal processing filter to an argument z-Nin the z-domain, being N said decimation ratio, i.e., the ratio between the first sample rate fs,i and the second sample rate fs,2. The delay operation by the specified amount may be implemented in the z-domain by using the argument z-N, however, the application in the z-do- main may be particularly advantageous if the signal processing filter for generating the first processed signal from the first intermediate signal is an infinite impulse response (HR) filter.

[0036] Preferably, in the analysis path, said initial filter coefficients cinmare calculated by means of a first shadow filter, said first shadow filter being a representation of the signal processing filter of the processing path. In the notation of the example used in equation (i), this means that for the first decimated signal d1 [n], the application of the first shadow filter with the initial coefficients cinmas

[0037] (iii) yields the same resulting signal as the first processed signal y1 [n] in the processing operation of the first intermediate signal x1 [n] displayed in equation (i) when using the first filter coefficients c1 j as defined via the initial filter coefficients cinmin equation (ii).

[0038] In an embodiment, for deriving the first output signal from the first processed signal, a low pass filter is applied to the first processed signal or to a signal derived from the first processed signal prior to said first output signal. The low pass filter helps to suppress undesired signal contributions well above the frequency range corresponding to the second sample rate fs,2 of the first decimated signal.

[0039] In an embodiment, said first input signal is generated as an analog signal, and said first intermediate signal is generated from said first input signal via an A / D conversion. This means in particular that the first intermediate signal is the first digital signal along the signal flow in the entire processing method. The A / D conversion generates the first intermediate signal from the first input signal at the first sample rate.

[0040] However, in an embodiment, the first input signal may also be given by a digital signal. In this case, the A / D conversion shall be considered as part of the generation of the first input signal from the environment sound, and the pre-processing step employed thereby. Said A / D conversion may occur at a higher sample rate fA / D than the first sample rate fs,i. Preferably, for generating the first intermediate signal, an initial decimation is applied to said first input signal or to a signal derived from the first input signal (i.e. , prior to the first intermediate signal, along the signal flow).

[0041] In an embodiment of particular advantage for said case, for deriving said first output signal from the first processed signal, an interpolation is applied to the first processed signal or to a signal derived from the first processed signal (i.e., prior to said first output signal, along the signal flow), said interpolation configured to compensate for said initial decimation. The interpolation in particular allows for “returning” to the sample rate prior to said initial decimation. A higher sample rate may be of advantage for the D / A conversion prior to generating the output sound signal via the output transducer, as it may reduce noise and / or artifacts.

[0042] In an embodiment, an acoustic feedback path from an output transducer of the hearing instrument to said first input transducer is estimated in the analysis path, and said estimation of the acoustic feedback path is based on the first processed signal and on the first decimated signal. This means in particular that for the suppression of an acoustic feedback, the estimation of the feedback path is based on the first decimated signal. Preferably, an adaptive filter for estimating said feedback path is adaptively controlled by using the first processed signal, and in particular, by using a decimated version of said first processed signal at the second sample rate (i.e., decimation of the first processed signal brings its signal contributions and its acoustic information to the second sample rate of the first decimated signal).

[0043] In a preferred embodiment, a compensation signal is generated in said processing path, using said estimation of the acoustic feedback path, and further using a delay in time-domain and / or a number of copies of a transfer function of a filter based on said estimation of the acoustic feedback path in frequency domain. In particular said delay and / or said number of copies of said transfer function for generating the compensation signal share the properties of the delay and the number of copies of said transfer function, respectively, for generating the first processed signal. Preferably, the compensation signal is generated by applying a filter based on said estimation of the acoustic feedback path to the first processed signal, or to a signal derived from the first processed signal.

[0044] In another expedient embodiment, said acoustic feedback path is estimated in the analysis path by means of a second shadow filter, said second shadow filter being a representation of an acoustic feedback path estimation of the processing path. In particular, said second shadow filter for estimating the acoustic feedback path shares the properties of the first shadow filter for calculating said initial filter coefficients.

[0045] The invention also discloses a hearing instrument, comprising at least a first input transducer configured to generate a first input signal from an environment sound, and a signal processing unit, wherein the hearing instrument is configured to perform the method described above. In particular, the signal processing unit is configured to execute all signal processing steps involved in said method.

[0046] The hearing instrument according to the invention shares the advantages of the method for low-delay signal processing in a hearing instrument according to the invention. Particular assets of the method and of its embodiments may be transferred, in an analogous way, to the hearing instrument and its embodiments, and vice versa.

[0047] The attributes and properties as well as the advantages of the invention which have been described above are now illustrated with help of drawings of embodiment examples. In detail, fig. 1 shows a schematic circuit diagram of a hearing instrument, fig 2 shows a signal flow chart for a hearing instrument according to the prior art, fig. 3 shows a signal flow chart for a hearing instrument according to an embodiment of the present invention, comprising an additional signal decimation in an analysis path, and fig. 4 shows a signal flow chart for a hearing instrument according to another embodiment of the present invention, comprising a feedback suppression filter.

[0048] Parts and variables corresponding to one another are provided with the same reference numerals in each case of occurrence for all figures.

[0049] In figure 1 , a schematic circuit diagram of a hearing instrument 1 is shown. The hearing instrument 1 is given by a hearing aid 2, which in the present case is embodied as a BTE device. However, the following considerations remain valid for other hearing aid designs (such as RITE, RIC or ITE), and also for more general hearing instruments not primarily intended and designed for providing support in case of a hearing loss.

[0050] The hearing instrument 1 comprises an acousto-electric first input transducer M1 , which in the present case is given by a microphone. The first input transducer M1 is configured to generate an electrical first input signal E1 from an environment sound 4. The first input signal E1 is processed in a signal processing unit 6, wherein a first output signal A1 is derived from the first input signal E1 by means of frequency-dependent signal processing, which in general may comprise frequency-dependent amplification and / or compression, as well as other signal processing techniques such as noise reduction or speech enhancement. Part of said signal processing for obtaining the first output signal A1 may be implemented by a corresponding filter in the signal processing unit 6, in a way yet to be described.

[0051] Finally, an electro-acoustic output transducer L1 generates an output sound signal 8 from the first output signal A1 . In the present context, an electro-acoustic output transducer shall be generally understood as any device which is intended, designed and configured to convert an electrical signal into a corresponding sound signal, whereby voltage and / or current fluctuations in the electrical signal are converted into corresponding amplitude fluctuations in the sound signal, and may in particular be given as a loudspeaker, or a so-called balanced metal case receiver, but also as a bone conduction receiver. Here, the output transducer L1 is preferably given by a loudspeaker positioned in an earpiece 9.

[0052] The hearing instrument 1 may comprise one or more additional input transducers (not shown in figure 1 ), each of which configured to generate a respective additional input signal from the environment sound. In this case, the one or more additional signals are also fed to the signal processing unit 6, along with the first input signal E1 , and the first output signal A1 is typically derived from both the first input signal E1 and the additional input signal(s). Furthermore, the hearing instrument 1 may be part of a hearing aid system (not shown), said hearing aid system possibly comprising another hearing instrument (i.e. , a binaural hearing system with two hearing aids, each of which configured to be worn by the user at one of his ears) and / or an external device connectable to the hearing instrument 1 (such as, e.g., a mobile phone for, inter alia, controlling the operation and / or certain functions of the hearing instrument 1 ). The validity neither of the concepts surrounding the invention as explained above nor of the considerations presented below are compromised by the presence of additional input signals and / or another hearing device or external device.

[0053] Figure 2 shows a possible signal flow for a hearing instrument similar to the one shown in figure 1 , according to the prior art. The hearing instrument 1 comprises the acousto-electric input transducer M1 (i.e. said microphone), an A / D converter (ADC) 12 and an initial decimator 14, as well as a signal processing filter 16, an interpolator 17, a D / A converter (DAC) 18, the electro-acoustical output transducer L1 (i.e. said loudspeaker). The first input transducer M1 generates the analog first input signal E1 from the environment sound 4, which is converted into a digital first input signal E1 [n] by the A / D converter 12, at a conversion sample rate fs,c, which may be typically in the order of 1 MHz to 5 MHz.

[0054] The digital first input signal E1 [n] is then decimated by the initial decimator 14, i.e., the digital first input signal E1 [n] is downsampled to a first sample rate fs,i < fs,c, and a low-pass filter is applied in order to spread out (or remove) quantization noise from the downsampling operation to higher frequency bands above the first sampling rate fs,i . The initial decimator 14 generates a first intermediate signal x1 as result of said decimation operation of the digital first input signal E1 [n], Due to the decimation, a delay is introduced into the processing of the digital first input signal E1 [n], said delay being of the order of the inverse bandwidth, i.e., ~ 1 / fs,i.

[0055] Next, for processing the first intermediate signal x1 (as part of the signal processing of the digital first input signal E1 [n]), the first intermediate signal x1 is split into a processing path 10 and an analysis path 20. The signal processing filter 16 is implemented in the processing path 10, and is applied to the first intermediate signal x1 , thereby generating a first processed signal y1 . The signal processing filter 16 is implemented as a time-varying low delay filter, obtaining its filter coefficients by means of an analysis of the first intermediate signal x1 performed by a signal analyzer 22 in the analysis path 20. In said signal analyzer 22, the first intermediate signal x1 may in particular be passed through an analysis filter bank (not shown) in order to divide the first intermediate signal x1 into a plurality of frequency bands, such that frequency-dependent signal processing parameters (not shown in detail in figure 2) such as frequency-dependent gains and / or frequency-dependent compression parameters (such as compression ratios or knee-points) for the first intermediate signal x1 may be determined. These frequency-dependent signal processing parameters in frequency domain may be transformed into time domain, obtaining the corresponding time-varying filter with respective first filter coefficients c1 j. These first filter coefficients c1j obtained in the analysis path 20 by means of an analysis of the first intermediate signal x1 are then used for the signal processing filter 16 in the processing path, i.e. , the signal processing filter 16 applied to the first intermediate signal x1 in the processing path 10 uses exactly the first filter coefficients c1 j obtained from the analysis of the first intermediate signal x1 in the analysis path.

[0056] The advantage of this implementation is that the signal processing filter 16 may be applied to the first intermediate signal x1 in the processing path 10 as a filter in time domain, i.e., an additional delay from a filter bank to divide the first intermediate signal x1 into respective frequency bands for the application of the signal processing can be avoided. Such a delay is then introduced only in the analysis path 20, such that, at worst, the first filter coefficients c1 j applied to the first intermediate signal x1 in the processing path 10 may carry an “intrinsic delay” in the sense that they are applied (in the processing path 10) to a portion of the first intermediate signal x1 which is a few milliseconds (i.e., the amount of delay introduced by the signal analyzer 22 in the analysis path 20) “shifted forward”, but the signal contributions of the first intermediate signal x1 themselves do not experience any extra “time-shift backwards” (i.e., a time delay) from this processing.

[0057] The first processed signal y1 generated in the processing path 10 by the application of the signal processing filter 16 onto the first intermediate signal x1 is then interpolated by means of the interpolator 17 in order to increase the sample rate of the first processed signal y1 , preferably back to the conversion sample rate fs,c, resulting in a digital first output signal A1 [n], Finally, after this interpolation step, the digital first output signal A1 [n] (at said increased sample rate) is converted into the analog first output signal A1 by means of the D / A converter 18, and said first output signal A1 is converted into the output sound signal 8 by means of the output transducer L1 .

[0058] Preferably, at least one out of and most preferably, the A / D converter 12 and the D / A converter 18 are based on sigma-delta modulation. Sigma-delta modulation is a generally advantageous method of conversion between analog and digital signals (and vice versa) , because it achieves high-resolution based on precise timing instead of having to rely on precisely-matched on-chip components. Likewise, the decimator 14 is based on a low-pass filter and an FIR filter.

[0059] The sigma-delta modulation technique may also be denoted oversampling or noise shaping. This is due to the fact that oversampling and noise shaping evoke the two basic principles of sigma-delta modulation. Oversampling by the A / D converter 12 spreads the quantization noise power over a bandwidth up to the conversion sample rate fs,c, which is much greater than the signal bandwidth of the fist intermediate signal actually used, and subsequently (related to and in the order of the fist sample rate fs,i ), the noise is shaped by providing that most of the noise energy (i.e. the quantization noise) is above said signal bandwidth (i.e. out-of-band). Consequently, the decimator 14 comprises a low-pass filter (not shown in detail in figure 1 ), and thus removes the out-of-band quantization noise and subsequently down sample the digital signal provided by the A / D converter 12 to the Nyquist rate.

[0060] For implementing the signal processing filter 16 with as little delay on the first intermediate signal x1 as possible, there is a tradeoff to consider. The initial decimation implemented by the decimator 14 (which, as mentioned, is preferably achieved via a sigma-delta modulation) introduces a delay in the order of the reciprocal of the resulting bandwidth, i.e., in the order of 1 / fs,i .

[0061] These delays could be reduced by increasing the first sample rate fs,i of the first intermediate signal x1 resulting from the decimator 14. If in an alternative scenario, the decimator 14 downsamples the digital first input signal E1 [n] to the first intermediate signal x1 having, e.g., a first sample rate of fs,i’ = k ■ fs,i with k e N, k > 1 , then (as explained above with help of equation (i)) the delay is reduced by a factor of k in the processing path, however, such a reduction comes at a cost of an increase in computational complexity (i.e., more calculation operations) and, thus, of power consumption in the order of 0(k2). Typical values of the first sample rate fs,i used for the signal to be processed in the processing path 10 (here: the first intermediate signal) are in the order of 10 to 60 kHz, preferably 16 to 34 kHz (corresponding to a delay in the range of 16 to 100 ps, preferably 30 to 65 ps only for the decimation, in case of a first order filter being used, and without even accounting for the interpolation). Thus, the present invention has a different approach for further reducing the delay.

[0062] Figure 3 shows a schematic signal flow for the hearing instrument displayed in figure 1 , according to an embodiment example of the present invention. Up to the generation of the first intermediate signal x1 , the signal flow shown in figure 3 resembles the one displayed in figure 2, i.e., the one known in the prior art. However, in the analysis path 20, another decimation step is applied to the first intermediate signal x1 by another decimator 24, generating thus a first decimated signal d1 from said first intermediate signal x1 , said first decimated signal d1 having a second sample rate fs,2 smaller than the first sample rate fs,i of the first intermediate signal x1 , i.e., fs,2 < fs,i .

[0063] Now, however, the signal analyzer calculates the first filter coefficients clj for the signal processing filter 16 in the processing path 10 in a different way: the signal analyzer 22 comprises a first shadow filter 26, for which respective initial filter coefficients cinmare calculated, and which is to be applied to the first decimated signal d1 , in the same or a similar way as, e.g., given by equation (iii). The initial filter coefficients cinmdetermined for application in the first shadow filter 26 onto the first decimated signal d1 (at the second sample rate fs,2) are now mapped to the first filter coefficients c1 j for the signal processing filter to be applied to the first intermediate signal x1 (at the first sample rate fs,i ) in the processing path 10. Said mapping is performed by means of a delay of m ■ (N - 1 ) depending on the decimation ratio N (i.e. , the ratio fs,i / fs,2 between the first and the second sample rates) for an m-th initial filter coefficient cinm. Thus, the m-th initial filter coefficient cinmis mapped to the first filter coefficient dm ■ N while other first filter coefficients c1 j for j m ■ N, the corresponding first filter coefficients c1 j are set to zero.

[0064] This delay operation yields a structure in the frequency domain in which the frequency window Qi corresponding to the first sample rate fs,i is filled with N copies of the transfer function H(o) of the first shadow filter 26 (defined over the frequency window Q2 corresponding to the second sample rate fs,2).

[0065] This way, in order to achieve a desired computational complexity by means of a sufficiently low sample rate (see discussion in the description of figure 2), the initial decimation performed by the decimator 14 does not have to result in this low sample rate paying the corresponding delay “penalty”. The first sample rate fs,i may then be, e.g., in the order of 200 kHz to 300 kHz (say, e.g., 256 kHz), or even higher, while the decimator 24 in the analysis path 20 applies a decimation to the first intermediate signal fs,i with a decimation ratio of, e.g., N = 8, giving rise to a second sample rate fs,2 in the order of 25 kHz to ~ 40 kHz for the first decimated signal d1 .

[0066] The additional delay corresponding to the bandwidth of the first decimated signal d1 at the lower second sample rate fs,2 is experienced only in the analysis path, i.e., the first filter coefficients c1 j applied to the first intermediate signal x1 via the signal processing filter 16 in the processing path 10, may be slightly delayed with respect to the “optimal” filtering coefficients (in the sense of an optimization algorithm performed on the underlying signal in order to find the optimal filter coefficients). However, the environment sound 4 (even in the presence of speech signals or other useful signals) is assumed to be sufficiently “inert” regarding the signal processing such that this delay is hardly perceivable. However, the delay would be sufficient to compromise certain signal quality standards when occurring to the signal contributions in the processing path 10. Figure 4 shows an alternative signal flow to the one displayed in figure 3, according to another embodiment example of the present invention. In the hearing instrument 1 given in figure 4, a feedback cancellation loop 30 is implemented, configured to suppress an acoustic feedback occurring along an acoustic feedback path 32 from the output transducer L1 towards the first input transducer M1 (which would couple in signal contributions already amplified in the signal processing unit 6 into the first input transducer M1 , leading to more and more amplification of these signal contributions in a possibly instable feedback loop).

[0067] The feedback cancellation loop 30 comprises a decimator 34 configured to apply a decimation to the first processed signal y1 , and a second shadow filter 36. The second shadow filter 36 is configured to estimate the acoustic feedback path 32 on the basis of the decimated first processed signal yT and the first decimated signal d1 . The estimation of said acoustic feedback path 32 preferably is performed by means of an adaptive filter (not shown), said adaptive filter being applied to the first decimated signal d1 and being controlled by the decimated first processed signal yT.

[0068] Preferably, in a similar way as in the first shadow filter 26 described above, second filter coefficients c2j are generated in the second shadow filter 36, said second filter coefficients c2j being used to parametrize a feedback cancellation filter 38 of the feedback cancellation loop 30. Preferably, the second filter coefficients c2j are generated from corresponding feedback suppression filter coefficients (not shown) of the second shadow filter by means of a delay (in time domain) and / or by means of a number of copies of a transfer function of the second shadow filter (in frequency domain) in a similar way to the generation of the first filter coefficients c1 j of the signal processing filter 16 in the processing path 10 from the initial filter coefficients cinmdetermined in the analysis path 20.

[0069] The feedback cancellation filter 38, using the second filter coefficients c2j, is applied in the feedback cancellation loop 30 to the first processed signal y1 (i.e. , the feedback cancellation loop 30 operates at the first sample rate fs,i , just as the processing path 10), generating a compensation signal u1 , which is subtracted at a subtraction node 40 from the output signal u2 of the initial decimator 14, generating thus the first intermediate signal x1 . Thus, in the embodiment shown in figure 4, the first intermediate signal x1 is not given directly by the output signal u2 of the initial decimator 14 (applied to the digital first input signal E1 [n]), but rather by the “error signal” generated from subtracting the compensation signal u1 generated in the feedback cancellation loop 30 from the “forward signal” (i.e. , the output signal u2 of the initial decimator 14).

[0070] Preferably, the decimation applied by the decimator 34 to the first processed signal y1 is the same as the decimation applied by the decimator 24 to the first intermediate signal x1 , i.e., the decimated first processed signal yT is preferably downsampled to the second sample rate fs,2. In the embodiment shown in figure 4, the generation of the first filter coefficients c1 j, preferably is analogous to the way shown with help of figure 3.

[0071] Even though the invention has been illustrated and described in detail with help of a preferred embodiment example, the invention is not restricted by this example. Other variations can be derived by a person skilled in the art without leaving the extent of protection of this invention.

[0072] Reference Numeral

[0073] 1 hearing instrument

[0074] 2 hearing aid

[0075] 4 environment sound

[0076] 6 signal processing unit

[0077] 8 output sound signal

[0078] 9 earpiece

[0079] 10 processing path

[0080] 12 A / D converter

[0081] 14 initial decimator

[0082] 16 signal processing filter

[0083] 17 interpolator

[0084] 18 D / A converter

[0085] 20 analysis path

[0086] 22 signal analyzer

[0087] 24 decimator

[0088] 26 first shadow filter

[0089] 30 feedback cancellation loop 32 acoustic feedback path 34 decimator

[0090] 36 second shadow filter

[0091] 38 feedback cancellation filter 40 subtraction node

[0092] A1 first output signal

[0093] A1 [n] digital first output signal c1j first filter coefficients c2j second filter coefficients cinminitial filter coefficients d1 first decimated signal

[0094] E1 first input signal

[0095] E1 [n] digital first input signal fs,c conversion sample rate fs,i first sample rate fs,2 second sample rate

[0096] H(o ) transfer function L1 output transducer

[0097] M1 first input transducer u1 compensation signal u2 output signal (of the initial decimator) x1 first intermediate signal y1 first processed signal y1 ’ decimated first processed signal

[0098] Q1 / 2 frequency window

Claims

Claims1 . A method for low-delay signal processing in a hearing instrument (1 ), wherein a first input signal (E1 ) is generated from an environment sound (4) by means of a first input transducer (M1 ) of the hearing instrument (1 ), wherein a digital first intermediate signal (x1 ) is generated from the first input signal (E1 ), said first intermediate signal (x1 ) having a first sample rate (fs,i), wherein in an analysis path (20), a decimation (24) is applied to the first intermediate signal (x1 ), resulting in a first decimated signal (d1 ) having a second sample rate (fs,2) smaller than the first sample rate (fs,i ), and by means of a signal analysis of the first decimated signal (d1 ), first filter coefficients (c1 j) of a signal processing filter (16) for the first intermediate signal (x1 ) are calculated, wherein in a processing path (10), the signal processing filter (16) using the first filter coefficients (c1 j) is applied to the first intermediate signal (x1 ) at the first sample rate (fs,i), generating a first processed signal (y1 ), and wherein a first output signal (A1 ) of the hearing instrument (1 ) is derived from the first processed signal (y1 ).

2. The method according to claim 1 , wherein in the analysis path (20), initial filter coefficients (cinm) of the signal processing filter (16) are calculated by means of the first decimated signal (d1 ), and wherein the first filter coefficients (c1 j) are derived from the initial filter coefficients (cinm) using a delay in time-domain and / or a number of copies of a transfer function (H(o)) of a filter based on the initial filter coefficients (cinm), said number being dependent on a ratio between the first sample rate (fs,i ) and the second sample rate (fs,2), in frequency domain.

3. The method according to claim 2, wherein said delay corresponds to the inverse of said second sample rate (fs,2), and / or wherein said number of copies of said transfer function (H(o)) corresponds to an application of the signal processing filter (16) to an argument z-Nin the z-domain, N being the ratio between the first sample rate (fs,i ) and the second sample rate (fs,2).

4. The method according to claim 2 or claim 3, wherein in the analysis path (20), said initial filter coefficients (cinm) are calculated by means of a first shadow filter (26), said first shadow filter (26) being a representation of the signal processing filter (16) of the processing path (10).

5. The method according to any of the preceding claims, wherein for deriving the first output signal (A1 ) from the first processed signal (y1 ), a low pass filter is applied to the first processed signal (y1 ) or to a signal derived from the first processed signal (y1 ).

6. The method according to any of the preceding claims, wherein said first input signal (E1 ) is generated as an analog signal, and said first intermediate signal (x1 ) is generated from said first input signal (E1 ) by means of an analog-to-digital conversion.

7. The method according to any of the preceding claims, wherein for generating the first intermediate signal (x1 ), an initial decimation (14) is applied to said first input signal (E1 ) or to a signal derived from the first input signal (E1 ).

8. The method according to claim 7, wherein for deriving said first output signal (A1 ) from the first processed signal (y1 ), an interpolation (17) is applied to the first processed signal (y1 ) orto a signal derived from the first processed signal (y1 ), said interpolation (17) preferably being configured to compensate for said initial decimation (14).

9. The method according to any of the preceding claims, wherein in the analysis path (20), an acoustic feedback path (32) from an output transducer (L1 ) of the hearing instrument (1 ) to said first input transducer (M1 ) is estimated, and wherein said estimation of the acoustic feedback path (32) is based on the first processed signal (y1 ) and on the first decimated signal (d1 ).

10. The method according to claim 9, wherein a compensation signal (u1 ) is generated in said processing path (10), using said estimation of the acoustic feedback path (32), and further using a delay in time-domain and / or a number of copies of a transfer function of a filter based on said estimation of the acoustic feedback path (32) in frequency domain.11 . The method according to claim 9 or claim 10, wherein in the analysis path (20), said acoustic feedback path (32) is estimated by means of a second shadow filter (36), said second shadow filter (36) being a representation of an acoustic feedback path estimation of the processing path (10).

12. A hearing instrument (1 ), comprising: a first input transducer (M1 ) configured to generate a first input signal (E1 ) from an environment sound (4), and a signal processing unit (6), wherein the hearing instrument (1 ) is configured to perform the method according to any of the preceding claims.

Images