EP2537351B1 - Method for the binaural left-right localization for hearing instruments - Google Patents

Description

The invention relates to a method and a system for improving the signal-to-noise ratio for output signals from a microphone arrangement of two or more microphones on the basis of useful acoustic signals occurring to the side of the microphone arrangement. Such a method and system can be used in hearing instruments, in particular in hearing aids that can be worn on the head of a hearing aid wearer. In this context, laterally is to be understood to mean in particular to the right and left of the head of the wearer of a binaural hearing aid arrangement.

Conventional directivity methods that have been used in hearing aids up to now offer the possibility of separating signals or noises that impinge on the hearing aid wearer from the front or the rear from the other ambient noises in order to increase speech intelligibility. However, they do not offer the possibility of emphasizing signals or noises from a source from the side that are incident from the left or right.

Previously known hearing aids only offer the possibility of emphasizing such lateral signals somewhat by transmitting the signal from the desired side to both ears. For this purpose, audio signals are transmitted from one side of the ear to the other and played there. As a result, however, the hearing aid wearer is presented with a mono signal, with the result that signal properties that make the localization of sound sources possible ('binaural cues') are lost. Such signal properties can be, for example, interaural level differences, i.e. that the level at the ear or hearing aid facing the noise or the signal source is higher than at the ear or hearing aid facing away.

The calculation of a conventional differential directional microphone is not a fully applicable solution, under Among other things, because signals with high frequency components cannot use a differential directional microphone without spatial ambiguities due to the so-called "spatial aliasing".

Such spatial ambiguities, i.e. The no longer unambiguous assignability of the spatial origin of a signal occurs when the right and left microphone signals of an acoustic original signal are subtracted from one another. The differential processing by subtracting the microphone signals normally allows a directional sensitivity of the microphone arrangement to be specified in a desired direction. If, however, the wavelength of the original acoustic signals is too small compared to the spatial distance between the microphones of the microphone arrangement, the spatial origin of an original signal can only be determined ambiguously or ambiguously.

In the document US 2003/147538 A1 describes a method for suppressing wind noise in output signals of a microphone arrangement composed of two microphones. For this purpose, power values and a coherence between the microphone signals are calculated for the individual microphone signals. By comparing the power values, it is recognized whether there is wind noise. In order to suppress a detected wind noise, a gain factor is calculated as a function of the cross-correlation and an estimate of the interference signal power.

In a scientific article ( JUNFENG LI ET AL. "Two-stage binaural speech enhancement with wiener filter based on equalization-cancellation model", APPLICATIONS OF SIGNAL PROCESSING TO AUDIO AND ACOUSTICS, 2009. WASPAS'09. IEEE WORKSHOP ON, IEEE, PISCATAWAY, NJ, USA, Oct 18, 2009, pp. 133-136 ) describes a binaural microphone arrangement, by means of which interfering noises in microphone signals can be suppressed on the basis of a Wiener filter. An interfering signal is estimated by differential processing of the microphone signals. An estimate of the signal-to-noise ratio is used to calculate the Wiener filter gain used on the basis of the "decision-directed approach".

In the document EP 1 465 456 A2 describes a binaural hearing device with a filter for noise suppression. A Wiener filter is used for noise suppression, the weighting factor of which is calculated from the cross-correlation between two microphone signals and the power of the individual microphone signals.

In a scientific article (" Microphone Arrays / Binarual Noise Suppression "In: James M. Kates:" Digital Hearing Aids ", January 1, 2008, Plural Publishing, San Diego ) describes several different filter techniques with regard to the reduction of background noise in microphone signals. One chapter describes microphone arrangements for the differential processing of a microphone signal. Another chapter describes the binaural noise reduction based on Wiener filtering. The binaural Wiener filter is based on the calculation of the cross-correlation between two microphone signals and the power of the individual microphone signals.

In the document EP 2 104 377 A2 describes a binaural hearing system with subband signal exchange. Low-frequency signal components are transmitted to the other hearing device, so that, for example, a beamforming algorithm can be implemented that uses signals from the other hearing device.

The object of the invention is to provide an improvement in the interfering signal-useful signal distance in acoustic signals taking into account a spatial direction of the signal source.

The invention according to claim 1 solves the problem in that it is viewed as a classic noise reduction problem. A binaural interference signal and a binaural useful signal are determined or estimated in the manner described below, which are used as input signals of a suitable filter, ie a Wiener filter, in which a gain factor is calculated and applied per frequency band, which is the same for both sides of the ear. By applying the same gain to both ears the interaural level differences are preserved, ie the localization of sounds or sound sources is made possible.

A basic idea of the invention is to process high and low frequency components (cutoff frequency in the range between 700 Hz and 1.5 kHz, for example approx. 1 kHz) differently. Filtering is carried out for low frequency ranges, ie Wiener filtering, based on differential preprocessing based on the computation of a differential binaural directional microphone, the preprocessing generating a left-hand and right-hand signal, usually with oppositely directed cardioid characteristics (kidney-shaped, directional sensitivity).

These two signals directed to the left and to the right on the basis of a conventional differential directional microphone are used as the basis for estimating the level of useful and interference noise from the side, these estimations in turn being used as input variables for the filtering, i.e. Wiener filtering can be used.

This filtering is then applied separately to each of the microphone signals of the microphone arrangement, and not to the common differential directional microphone signal of the binaural arrangement, which was calculated as the output signal of the conventional directional microphone.

The advantage e.g. Compared to the use of omni signals, greater differences between the left and right side are artificially generated by the upstream directional effect, which are expressed in an increased noise suppression of signals arriving from the direction to be suppressed.

The invention provides for a pre-filtering based on the calculation of a conventional differential directional microphone and subsequent filtering, i. E.

Perform Wiener filtering as explained above and use the natural head shadowing effect as a pre-filter for noise and useful noise estimation for subsequent Wiener filtering in high frequency ranges (cutoff frequency in the range between 700 Hz and 1.5 kHz, e.g. approx. 1 kHz).

The determination of the interference and useful noise estimate using the head shadowing effect is carried out as follows: The monaural signal facing the desired side is used as the useful signal estimate, that of the opposite side as the interference signal estimate. This is possible because, especially at higher frequencies (> 700 Hz or> 1 kHz), the head shadowing effect causes considerable attenuation of the signal on the opposite side.

These two signals directed to the left and to the right on the basis of a signal pre-filtered by head shadowing are used as the basis for estimating the level of useful and interfering noise, and these estimations are in turn used as input variables for the filtering, preferably Wiener filtering .

This filtering is then applied separately to each of the microphone signals of the microphone arrangement.

The advantage e.g. Compared to the use of omni signals, greater differences between the left and right side are artificially generated by the upstream directional effect, which are expressed in an increased noise suppression of signals arriving from the direction to be suppressed.

The respective preprocessing generates a signal directed to the left and a signal directed to the right for the low and high frequency ranges, usually with oppositely directed cardioid characteristics (kidney-shaped, direction-dependent sensitivity). These respective directional signals are used as a basis for estimating the respective lateral useful and background noise levels. The respective useful and interfering sound levels are in turn used as input variables for the filtering, preferably Wiener filtering. By combining the respective filtering process for high and low frequency ranges filtering over the entire frequency range can thus be achieved.

According to the invention, the acoustic signals are broken down into frequency bands, and the filtering, i. Wiener filtering, made specifically for each of the frequency bands.

In a further advantageous development, the filtering, i. Wiener filtering, made depending on the direction. The directional filtering can be carried out in a conventional manner.

Fig 1:

Pegel des linkseitigen und rechtsseitigen Mikrofons für ein umlaufendes Signal bei 1 kHz

Fig 2:

Richtungsabhängig gedämpftes Signal bei 1 kHz nach Anwendung Wiener-Filter für das linksseitige und rechtsseitige Mikrofon

Fig 3:

Gerichtetes differenzielles Richtmikrofonsignal sowie jeweiliges Wiener-vorgefiltertes Mikrofonsignal für Frequenzen von 250 Hz und 500 Hz nach links (bei 270°)

Fig 4:

Schematische Darstellung des Verfahrens zur Verbesserung des Signal-Rausch-Abstands bei binauraler Seitenwahrnehmung

One or more of the following parameter values is advantageously determined or estimated as the useful signal level and / or as the interference signal level: energy, power, amplitude, smoothed amplitude, averaged amplitude, level. Further advantageous developments and advantages can be found in the dependent claims and the following figures, including the description. Show it:

Fig 1:

Levels of the left and right microphone for a circulating signal at 1 kHz

Fig 2:

Directional attenuated signal at 1 kHz after using Wiener filter for the left and right microphone

Fig 3:

Directional differential directional microphone signal as well as the respective Wiener pre-filtered microphone signal for frequencies of 250 Hz and 500 Hz to the left (at 270 °)

Fig 4:

Schematic representation of the method for improving the signal-to-noise ratio in binaural side perception

In Figure 1 are the levels of the hearing aid microphones or - microphone arrangements of the left (provided with the reference symbol L2 in the figure) and right (reference symbol L1) side of the ear a binaural hearing aid arrangement for a circumferential signal, ie for a signal source positioned in the illustrated circumferential spatial directions, shown at 1 kHz. A difference of 6-10 dB can be seen, ie the level L2 of the left-hand microphone or microphone arrangement is 6-10 dB higher for a left-hand signal (270 °) than the level L1 of the right-hand microphone or microphone arrangement; this level difference increases at higher frequencies.

If now e.g. If you want to hear to the left (270 °), the right signal L1 is used as an interfering sound signal, the left L2 as a useful sound signal. On the basis of this background noise and useful noise signal, the input variables for filtering, e.g. a Wiener filtering, can be estimated.

For Wiener filtering, respective useful signal and interference signal levels are determined or estimated from the useful signal and interference signal. These were used as input variables for a Wiener filter, i.e.: $$\mathrm{Wiener\; filter}=\mathrm{Useful\; signal\; level}/\left(\mathrm{Useful\; signal\; level}+\mathrm{Noise\; level}\right)$$

In Figure 2 the direction-dependent attenuation is shown, which results from the application of the Wiener formula for a circumferential (360 °) signal at 1 kHz. The result is the directionally attenuated signal L4 for the left-hand microphone or microphone arrangement and L3 for the right-hand microphone or microphone arrangement.

In comparison with the previous figure, it can be seen that the interaural level differences are retained. Signals from the right are viewed as interfering signals and are reduced, signals from the left remain undamped. The spatial impression, ie the signal information from where the signals come from, is retained because the level differences are retained. Hit signals from both sides on, there is a reduction depending on the ratio of useful and background noise estimate according to the well-known Wiener formula.

As described above, it is proposed to make use of the natural head shadowing effect in order to use the signals pre-filtered by the head shadowing effect as interference and useful sound signals for determining the input variables of a filter on a filter, e.g. Wiener Filter, based on the noise suppression approach. Since the head shadowing effect is particularly pronounced at high frequencies (> 700 Hz or> 1 kHz), but decreases further and further towards lower frequencies, this method can advantageously be used particularly for frequencies above 1 kHz.

For low frequencies (<1.5 kHz or <1 kHz), the solution explained above cannot be used optimally because the head shadowing effect is too low. In the low frequency ranges, the method described below can also be used, which can also be used separately and exclusively.

Since for low frequencies (<1.5 kHz or <1 kHz) it applies that the binaural microphone distance at the head of a hearing aid wearer is small enough compared to the wavelength, no spatial ambiguities ('spatial aliasing') arise. Therefore, at low frequencies (<1.5 kHz or <1 kHz) of the original acoustic signal with the microphone arrangement of a left-hand and a right-hand microphone or microphone arrangement on the head of a hearing aid wearer, a conventional differential directional microphone that "looks" or "looks" to the side can be used. hears ", be calculated.

The output signal of such a directional microphone could simply be used directly to generate a lateral directional effect at low frequencies. The directional signal determined in this way could then be reproduced identically on both ears or hearing aids of the hearing aid wearer. However, this would have the consequence that the localization ability in this frequency range would be lost, since only one common output signal would be generated and presented for both sides of the ear.

Instead, both a left and a right-pointing signal are calculated based on a conventional directional microphone, and these signals are used as interference or useful sound signals for subsequent filtering, preferably with a Wiener filter, depending on the desired direction of the useful signal. This filter is then applied separately to each of the microphone signals of the microphone arrangement, and not to the common directional microphone signal calculated as the output signal of the conventional directional microphone.

In Figure 3 shows the effect of the previously explained audio signal processing in low frequency ranges. For this purpose, for frequencies of 250 Hz L8 and 500 Hz L5, a left (at 270 °) left directed "hearing" or "seeing" was calculated. As part of the pre-filtering, a conventional differential directional microphone signal directed to the left was initially calculated as the useful signal and a signal directed to the right was calculated as the interference signal (solid lines in the figure). The directional microphone signals have the usual kidney / anti-kidney-shaped (cardioid / anticardioid, also: card / anticard) direction-dependent sensitivity characteristics.

The useful signal and interference signal levels were determined or estimated from the useful signal and interference signal. These were used as input variables for a Wiener filter, i.e.: $$\mathrm{Wiener\; filter}=\mathrm{Useful\; signal\; level}/\left(\mathrm{Useful\; signal\; level}+\mathrm{Noise\; level}\right)$$

Such a Wiener filter was calculated for each frequency range (ie 250 Hz and 500 Hz in the figure) for all spatial directions and applied individually to each of the directional microphone signals. This results in a Wiener-prefiltered, direction-dependent sensitivity characteristic for each of the directional microphone signals, which are shown in the figure by dashed lines L6 and L7.

It can be seen in the figure that a higher attenuation is achieved in the interfering signal direction (ie right, 90 °) than in the useful signal direction (ie left, 270 °). It can also be seen that the level differences are largely retained (namely a higher level of the left L7 compared to the right microphone signal L6) and thus a spatial assignment of the original acoustic signal remains possible for the hearing aid wearer.

The filter methods described above for high and low frequency ranges can e.g. can be used individually for high or low frequencies in hearing instruments to be worn on the head. However, they can also be used in combination and complement one another in a particularly advantageous manner over the entire frequency range of a hearing instrument to be worn on the head.

In Figure 4 the previously explained method for improving the signal-to-noise ratio in binaural side perception is shown schematically.

In step S1, a binaural microphone arrangement picks up acoustic signals. Such a microphone arrangement comprises at least two microphones, one each to be worn on the left or right side of the head of a hearing aid wearer. The respective microphone arrangement can also each comprise a plurality of microphones, which can, for example, enable a directional effect for perception to the front and back.

In step S2, a lateral direction is determined in which the highest sensitivity of the microphone arrangement should be directed. The direction can for example be automatically dependent on an acoustic analysis of the ambient noise or can be determined depending on a user input. The direction in which the greatest sensitivity is chosen is that direction in space in which the source of the useful acoustic signals is or is presumably located. In the present case, it is therefore also referred to as the useful signal direction. The microphone or microphone arrangement located in this direction is analogously also referred to here as a useful signal microphone.

In step S3, analogously to the step explained above, a lateral direction is determined in which the lowest sensitivity of the microphone arrangement should be directed. In the present case it is therefore also referred to as the interfering signal direction and the microphone or microphone arrangement located in this direction is referred to as the interfering signal microphone.

In step S4, the output signals of the microphones are broken down into a frequency range with high frequencies above a cutoff frequency of at least 700 Hz, possibly also 1 kHz, and a frequency range with low frequencies below a cutoff frequency of 1.5 kHz, possibly also 1 kHz.

In steps S5 to S7, the microphone signals are processed further in the high frequency range. In step S5, a useful signal level is determined or estimated as a function of the output signal of the useful signal microphone.

In step S6, an interference signal level is determined or estimated as a function of the output signal of the interference signal microphone.

In step S6, a filter, preferably a Wiener filter, is calculated using the useful signal level and interference signal level determined above. The signal levels and the filtering can be determined for the entire high frequency range. However, it can also be broken down into frequency bands within the high frequency range and the filtering can be done individually for each of the frequency bands.

In step S7, the previously calculated filter is applied separately to the respective output signals of the right-hand and left-hand microphones or microphone arrangements in the high frequency range.

In steps S8 to S13, the microphone signals of the low frequency range are processed further. In step S8, a conventional differential binaural directional microphone with high sensitivity in the direction of the useful signal is calculated, whereby a second useful signal is obtained.

In step S9, a conventional differential binaural directional microphone with high sensitivity in the interfering signal direction is calculated, whereby a second interfering signal is obtained.

In step S10, a second useful signal level is determined or estimated as a function of the second useful signal.

In step S11, a second interference signal level is determined or estimated as a function of the second interference signal.

In step S12, a second filter, preferably a Wiener filter, is calculated using the previously determined second useful signal level and second interference signal level. The second signal levels and the filtering can be determined for the entire low frequency range. However, it can also be broken down into frequency bands within the low frequency range and the filtering can be carried out individually for each of the frequency bands.

In step S13, the previously calculated filter is applied separately to the respective output signals of the right-hand and left-hand microphones or microphone arrangements in the low frequency range.

In step S14, the filtered output signals of the microphones of both frequency ranges or, in the case of further breakdown into frequency bands of all frequency bands, are combined to form a filtered output signal of the binaural microphone arrangement. $$\mathrm{In}\mathrm{one}\mathrm{further}\mathrm{further\; education}\mathrm{becomes}\mathrm{of\; the}\mathrm{Gain\; factor}\left(\mathrm{Viennese}\right)\mathrm{certainly}\mathrm{according\; to}\mathrm{of\; the}\mathrm{formula}\mathrm{Gain\; factor}\left(\mathrm{Viennese}\right)=\mathrm{Useful\; signal\; level}/\left(\mathrm{Useful\; signal\; level}+\mathrm{Noise\; level}\right).$$

In a further development, the useful signal microphone is arranged on a hearing aid wearer on the right and the interference signal microphone on a hearing aid to be worn on the left or vice versa.

In a further development, one or more of the following is estimated as the useful signal level and / or as the interference signal level: energy, power, amplitude, smoothed amplitude, averaged amplitude, level.

Claims (5)

1. Method for improving the signal-to-noise difference for laterally impinging acoustic useful signals, comprising the steps:

   - Recording (S1) of acoustic signals by a right-side microphone and a left-side microphone of a binaural microphone arrangement,

   - Definition of a first relevant frequency range and a second relevant frequency range located at higher frequencies with respect to first relevant frequency range, wherein a limit frequency between the first relevant frequency range and the second relevant frequency range is chosen in the range between 700 Hz and 1.5 kHz,

   - Definition (S2, S3) of a spatial direction as useful signal direction and a spatial direction as noise signal direction,

   - Determination (S9) of a binaural noise signal in the first relevant frequency range by means of differential preprocessing of monaural output signals (L1, L2) of the binaural microphone arrangement, for which in the useful signal direction, a lower sensitivity than in the noise signal direction is achieved,

   - Determination (S8) of a binaural useful signal in the first relevant frequency range by means of differential preprocessing of the monaural output signals (L1, L2) of the binaural microphone arrangement, for which in the useful signal direction, a higher sensitivity of the binaural microphone arrangement than in the noise signal direction is achieved,

   - Determination (S 11) of a first noise signal level in the first relevant frequency range in dependence of the binaural noise signal,

   - Determination (S10) of a first useful signal level in the first relevant frequency range in dependence of the binaural useful signal,

   - Definition (S2, S3) of the microphone of the binaural microphone arrangement which is located closer to the source as the useful signal microphone, and of the microphone of the binaural microphone arrangement which is located more remotely as the noise signal microphone,

   - Determination (S6) of a monaural second noise signal level in the second relevant frequency range in dependence of the monaural output signal (L1) of the noise signal microphone,

   - Determination (S5) of a monaural second useful signal level in the second relevant frequency range in dependence of the binaural output signal (L2) of the useful signal microphone,

   - Ascertainment (S12) of a first gain factor for the amplification of the acoustic signals being recorded with the microphones in the first relevant frequency range, by means of a first Wiener filter in dependence of the first noise signal level and the second noise signal level, and of a second gain factor for the amplification of the acoustic signals be recorded with the microphones in the second relevant frequency range, by means of a second Wiener filter in dependence of the second noise signal level and the second useful signal level, and

   - Application (S7, S13) of the first gain factor in the first relevant frequency range and the second gain factor in the second relevant frequency range separately to the respective output signals (L1, L2) of the right-side microphone and the left-side microphone of the binaural microphone arrangement.
2. Method according to claim one, comprising further the step

   - Direction-dependent ascertainment of the first gain factor.
3. The method according to one of the preceding claims, wherein the first and/or second gain factor is ascertained according to the formula gain factor = useful signal level / (useful signal level + noise signal level).
4. The method according to one of the preceding claims, wherein the useful signal microphone is arranged at a hearing aid to be worn by a hearing aid user at his right side, and the noise signal microphone is arranged at a hearing aid to be worn by a hearing aid user at his left side, or vice versa.
5. The method according to one of the preceding claims, wherein as a useful signal level and/or a noise signal level, one or more out of the following parameter values are determined:
   Energy, power, amplitude, smoothest amplitude, average amplitude, level.

Images