Method and device for noise suppression in a redundant acoustic signal

0001The invention relates to a process for the Eliminating glitches from a redundant acoustic input signal an acoustic reproducing apparatus, such as a mobile telephone or a hearing aid, in which the disturbance in a Sub-frequency range of an overall frequency range of is the input signal and concentrated on a device Performing the method.

0002At a reception of acoustic signals, such as speech, is in interference-prone environments, for example in a Motor vehicle, in an industrial setting, in which Machine noise occurs, and in public areas, the Problem to perform an elimination of interference of the speech signal, before it by an acoustic reproduction device is reproduced. Here are unwanted Interference signal of the input signal as much as possible be lowered. Also occurring artifacts are avoided are, even those relating to the interference signal.

0003For eliminating glitches from acoustic signals, it is known, so-called. use gradient microphones. These have a Directivity, which is used to Interference signal from directions other than the Microphone main direction towards a desired signal in attenuate microphone main direction. These microphones have However, the disadvantage that they have an increased sensitivity against wind and breath noise show. These Sensitivity through the use of a windbreak be improved, but which is dimensioned so large that he did not in acoustic reproduction means any establishment can be used. For example, the use of a windscreen with a mobile phone or a hearing aid as acoustic reproduction devices not common.

0004In hearing aids gradient microphones are regularly on the Base two omnidirectional microphones realized, wherein a so-called. "delay & substract" approach is used. The However, directivity of this approach is in particular in the lower frequency range of the input signal is very sensitive against an unavoidable in practice Misalignment of the microphone characteristics. Also give difficulties relating to the internal Microphone noise.

0005To suppress noise in acoustic signals is also the use of statistical Störgeräuschunterdrückungsverfahren known as spectral Subtraction method, the Wiener filter, method of Ephraim & Malah, R. Martin, etc. These methods use the different statistical characteristics of voice payloads and interference signals from. It is assumed, that a spurious signal is significantly stationary as a voice payload. The interference signal is typically in estimated speech pauses and continuously frequency specific of the disturbed input signal subtracted. The estimate of the Interference component is particularly at a transient disturbance difficult, especially when the interfering signal has a higher Level than the useful signal.

0006When subtracting the estimated noise and interference component occur also artifacts that the usual way as "Musical Tones "are referred to. The remaining Reststörsignal after carrying out the Störgeräuschunterdrückungsverfahrens often has no longer the original Störsignalcharakter on, ie the noise sounds different, which for many users acoustic reproduction devices unfamiliar and disturbing is. At low signal-to-noise ratio, For example, <5 dB, denied the Substraktionsprinzip in usually so that the disturbed input signal by the user usually judged as sound better is as the processed signal of after implementation Störgeräuschunterdrückungsverfahrens results.

0007Proceeding from this, the invention has the object, a method for eliminating glitches from a redundant acoustic An acoustic signal reproducing apparatus as well as a to provide apparatus for carrying it out, in which the sound quality of the resulting acoustic is output improved.

0008This object is achieved in terms of the process by a method for eliminating glitches from a redundant acoustic Input signal of an acoustic reproducing apparatus, such as a mobile phone or a hearing aid, wherein the disorder in a sub-frequency range of an overall frequency range of is the input signal is concentrated, characterized that the method comprises the sequential steps of:<sl><li>a) removing the part of the frequency range of the input signal, in which the disturbance is concentrated,</li><li>b) dividing the intensity of the in step a) the transmitted frequency range of the input signal into a be retained input part and a further processing input part,</li><li>c) synthesizing the remote in step a) Sub-frequency range of the input signal due to the further to be processed input signal portion and</li><li>d) combining the input signal to be retained portion from step b) and the synthesized input signal portion from step c) for bringing about a relative to the Input signal interference reduced output signal.</li></sl>

0009A major difference of the process towards the Method according to the prior art, it is the Sub-frequency range of the input signal, in which a disturbance is concentrated to completely discard. In contrast, was trodden always the way in the prior art, the edit relevant sub-frequency range so that a Interference is minimized.

0010After the procedure, the remote part of the frequency range Input signal, in which the disturbance is concentrated, synthesized, and due to an unremoved Frequency range of the input signal. is at this point utilizes the fact that the not distant or further processing input part due to the redundancy of the acoustic input signal conclusions about the Frequency response of the remote part of the frequency range allowed. For example, be assumed that, for each acoustic sound an assigned frequency spectrum exists, wherein a portion of frequency spectrum may be used to a on the part of the frequency spectrum adjacent spectrum . synthesize

0011After the procedure is the wish to retain Input part, ie in not substantially Interference-prone part of the total frequency range of Input signal, with the synthesized part of the frequency range, in which the disturbance falls merged to one against the input signal failure reduced output signal produce.

0012Preferably, in step c) synthesizing the Sub-frequency range of the input signal by means of a Bandwidth expansion procedure performed. For example is an expansion of the bandwidth in the lower frequency range (&lt;300 Hz) from the EP 0994464 A a recovery of Signal portions of the lower frequency range of a by a High pass towards lower frequencies limited Known speech signal, wherein the described High-pass filtering, for example, in voice transmission is performed via a telephone at the far.

0013The restoration is carried out by generating Frequencies of the lower frequency range by a non-linear signal processing, by means of the subharmonic Frequencies of Sig Nals generated and added to the high pass are added. It should be noted that in EP 0 994 464 A described method is not the elimination of interference of at a mobile phone or a hearing aid incoming signal has to the object. In contrast, the new method among others, mobile phones, hands-free, Hearing aids and other communication terminals for a mobile or use in interference-prone environments are provided, can be used.

0014Preferably, in step c), the synthesized Sub-frequency range of the input signal for adapting to maintaining the input signal portion of the input signal filtered. This ensures that the in d) held merge these two signal parts a trouble-reduced output signal results, which Frequency spectrum as well as possible the Total frequency spectrum of the useful signal of the input signal corresponds to her in particular on the intensity. Of course, must in synthesizing the in step a) filtered and remote part of the frequency range Input signal, the characteristics of the in step a) Spectral filter employed are taken into account, because they determine the frequency band to be synthesized. Special effects are obtained for the edge area a filter used.

0015For many environmental characteristics, such as a Mobile phone or a hearing aid, it is convenient that the in Step c) synthesized part of the frequency range Input signal on the low frequency side of the is the total frequency range of the input signal. For example, there are disturbances in a motor vehicle or called. pop and flow noises caused by wind and breath typical Way in the low frequency range of the input signal. Frequently are the intensities of interference much higher than the Nutzsignalintensitäten. When trying these disorders with the above-mentioned spectral subtraction of the prior eliminate art, this can only be sold done of artifacts that are present at high intensity. The new method does not exhibit these limitations.

0016It is also possible, with the process to disturbances reduce or avoid the high frequency on the Side of the total frequency range of the input signal are. Also disorders of in an intermediate frequency range concentrated entire frequency range of the input signal are, can be suppressed.

0017Another advantage of this new process consists in that also strongly unsteady disturbances may be removed, which is not possible with spectral subtraction method. The new method can favorably even with a Gradientenmikrofon be combined because its drawbacks, such as pop and flow noise sensitivity compensated, can.

0018The above object is in terms of a device for Carrying out the method achieved by a device for Play or forwarding a redundant acoustic Input signal, such as a mobile phone or a hearing aid, the Störunterdrückungsmittel to suppress in a Sub-frequency range of the total frequency range of Input signal of the input signal disturbance concentrated , characterized, that Störunterdrückungsmittel comprise:<ul><li>a spectral filter for removing the Sub-frequency range of the input signal in which the interference is concentrated,</li><li>a filter downstream division means for Dividing the intensity of the transmitted frequency range of the input signal into a to be retained Input part and a further processed Input part,</li><li>a dividing device downstream Processing means of the further processing Signal portion of the input signal is fed, for Synthesizing the filtered part of the frequency range Input signal due to the further processed Input part and</li><li>an output signal generating means, the be retained input part of the Dividing means and the synthesized Input signal part of the processing means are fed, to be retained for combining the Input signal portion and the synthesized Input part for bringing about a relative to the Input signal interference reduced output signal.</li></ul>

0019The operation of this device is clear from the above-discussed new methods, because it provides the required for the individual process steps Facilities ready.

0020It is to be considered as preferred that the apparatus includes a Means for detecting whether the input signal in the Stop band of the filter detects an anomaly has, and a Switch for bridging the Störunterdrückungsmittel due to a lack of disturbance in the stopband of the filter is provided. In this way it is ensured that the Störunterdrückungsmittel only be switched on when actually a disturbance in the stop band of the filter is.

0021It is also regarded as advantageous that the Apparatus comprising means for detecting or estimating a having intensity of the disturbance and a switch for Bridging the Störunterdrückungsmittel due to the Intensity detection / estimation is provided. This allows it that the Störunterdrückungsmittel only switched be when deemed appropriate due to the intensity of the disturbance is felt necessary. Here is a manual Setting a threshold for the detection or Estimating the intensity of the disturbance by a user For example, a mobile phone possible. Instead of Switch for bridging the Störunterdrückungsmittel can preferably also includes means for blending the Input signal portion in the stop band of the filter and synthesized signal portion, depending on the degree of disturbance in Be provided stopband of the filter.

0022both for the method and for the device emphasized that a failure in principle in a any part of the frequency range of a total frequency range may be concentrated an acoustic signal. The respective removed or filtered out sub-frequency range is then preferably by means of a bandwidth expansion procedure and synthesized with the retained frequency range was not filtered, brought together to a trouble-reduced output signal produce.

0023The invention will now by way of Example, with reference to the drawing explained in more detail. The drawing shows a block diagram a device for eliminating glitches from a redundant acoustic input signal.

0024The single drawing figure shows a procedure and for performing the process required for the components enhance the sound of an afflicted with a disorder redundant acoustic input signal 1, such as a Speech signal. Speech signals, it is inherent that the material to any given time present no frequency spectrum is completely needed to make a statement about meet, which, according to the current frequency spectrum assign is. Insofar voice signals are redundant, ie to a part of the frequency spectrum can be dispensed with, namely substantially without loss of information the overall frequency spectrum. This is because from the retained on the frequency spectrum part of the frequency spectrum, was dispensed with, conclusions can be drawn.

0025In particular and as can be seen in the drawing, 1 the input signal, after which it from a receiving module for example, receive a mobile phone or hearing aid was fed to a spectral high-pass filter 2 having a has defined filter characteristics that the present Embodiment frequencies above 300 Hz in Essentially lets pass without attenuation, while frequencies are strongly attenuated below 300 Hz. The value of 300 Hz thus forms the cut-off frequency f_A the high pass filter. 2

0026From the high-pass filter 2 of the signal reaches the the transmitted frequency range to a branching point 3 as dividing means, which is in a beizubehaltendes subsignal 4 and in a further processed partial signal is 5 divided.

0027Both the signal to be retained and the further 4 processed signal 5 having essentially no Frequency components in the range of below 300 Hz, so that a Disorder that is concentrated below 300 Hz, from the is the input signal 1 removed. In this respect refers to the embodiment explained a disorder responsive to the low-frequency side of the entire frequency range of the is the input signal 1 is concentrated.

0028The next step consists in the fact that the High-pass filter 2 remote part frequency spectrum to synthesize, and with the aid of the transmitted Frequency spectrum of the signal to be processed further. 5

0029The next signal to be processed 5 is first a Low-pass filter 6 fed that in the illustrated Embodiment frequencies higher than 2 kHz strongly attenuates. The low pass filter 6 is for carrying out the Method not mandatory required, simplified but a signal processing in a non-linear Processing device 7. In particular, by the Use of non-linear processing means 7 called. Intermodulation effects occur, their appearance by low-pass filter 6 is reduced or avoided.

0030The non-linear processing means 7 generates now either directly from the further processed signal 5 or from an output signal of the low-pass filter 6 a low frequency spectrum, and taking into account the Characteristic of the high-pass filter 2, which is already known and directly in the non-linear signal processing is considered. The non-linear Processing means 7 is used for recovery of Amplitudes of the fundamental frequency and / or missing harmonics in Depending on the underlying speech signal.

0031The non-linear processing means 7 leaving Signal thus comprises Ideally, the fundamental frequency and the missing harmonics of the input signal 1, the in High-pass filter 2, have been together with the disorder, removed are.

0032The output signal in the form of a synthesized Speech signal of the nonlinear processing means 7 is then fed to a band-pass filter 8, the frequencies below the cut-off frequency f_A 300 Hz high pass filter 2 and a lower limit frequency f_B for example, 100 Hz passes. Basically, the bandpass filter 8 can also by a low pass filter with a cutoff frequency of 300 Hz be replaced. However, when DC components from the Output of nonlinear processing means 7 are to be removed should be a band-pass filter 8 with the Appropriate filtering is performed.

0033An output signal of the bandpass filter 8 is a Node 9 as an output signal generation means fed at which it for the bandwidth extension to the be retained control unit 4 is added to a to obtain trouble-reduced output signal 10, Ideally, all voice-useful signal frequencies of should have input signal 1 with the right amplitude.

0034The illustrated in the drawing apparatus Liberation of the input signal 1 of a fault can also be modified in so far that a high frequency on the Side of the frequency spectrum of the input signal lying 1 Interference is removed by a low pass filter, in which case a nonlinear processing unit would provide that a bandwidth extension method to high frequencies can perform. In case of high frequency disturbance would the low-pass filter 6 from the figure by a suitable high-pass filter to replace. Also, the band-pass filter 8 from the figure would be by exchange a filter having a at the junction point 9 suitable linking the synthesized signal portion with the be retained signal part 4 permits.

0035The disorder may also within the entire frequency range the input signal 1 are. In this case, either of lying on the high frequency side of the fault Frequency range of the input signal on the one or the low-frequency side of the fault lying frequency range of 1 for performing an input signal Bandwidth expansion procedure can be utilized wherein the non-linear processing unit for closing a Frequency gap is that of through a bandpass filter instead High-pass filter 2 is induced.

0036In another, not shown in the drawing Embodiments of the invention, means may be provided be that, depending on external parameters, a switching the apparatus according to the drawing causes. Examples may be mentioned a means for detecting whether the Input signal in the stopband of the highpass filter 2 a Disorder or shows a device for detecting or Estimating an intensity of the disturbance. in both cases the respective means for actuating a switch which is arranged before the high-pass filter 2 is formed.

0037It is considered particularly advantageous if instead of a hard switch means for blending the Input signal 1 in the stopband of the respectively used Filter with the synthesized signal part is provided. The degree of transition, one of the intensity be present depending disorder. Such a device can for controlling a transition of the type mentioned for example be provided at the junction point 9, wherein the device is a measure of the estimate or is supplied Intensiät an interference signal as a control variable.