JP5150165B2 - Method and system for providing an acoustic signal with extended bandwidth - Google Patents

Abstract

The invention is directed to a method for providing an acoustic signal with extended bandwidth, comprising automatically determining a current upper and a current lower bandwidth limit of a received acoustic signal, automatically determining at least one complementary signal to complement the received acoustic signal between a predefined lower broadband bandwidth limit and the current lower bandwidth limit and/or between the current upper bandwidth limit and a predefined upper broadband bandwidth limit, wherein the predefined lower broadband bandwidth limit is smaller than the current bandwidth limit and the predefined upper broadband bandwidth limit is larger than the current upper bandwidth limit, automatically assembling the at least one complementary signal and the received acoustic signal to obtain an acoustic signal with extended bandwidth.

Description

  The present invention is directed to a method and system for providing an acoustic signal, particularly a speech signal, having an extended bandwidth.

  An acoustic signal transmitted over an analog or digital signal path typically suffers from the obstacle that the transmitted acoustic signal is very different from the original signal because the signal path has only a limited bandwidth. For example, in the case of a conventional telephone connection, a sampling rate of 8 kHz is used, resulting in a maximum signal bandwidth of 4 kHz. Compared to audio CDs, speech and audio quality are greatly reduced.

  In addition, many types of transmission exhibit additional bandwidth limitations. For analog telephone connections, only frequencies between 300 Hz and 3.4 kHz are transmitted. As a result, only a 3.1 kHz bandwidth is available.

  In principle, the bandwidth of a telephone connection can be increased by using wideband or wideband digital coding and decoding methods (so-called wideband codecs). However, in such cases, both the sender and receiver need to support corresponding coding and decoding methods, which require the implementation of new standards.

  As an alternative, a system for bandwidth extension may be used, for example, as described in [1] or [2]. These systems should be implemented only on the recipient side so that existing telephone connections do not need to be changed. In these systems, the missing frequency component of an input signal having a small bandwidth is estimated and added to the input signal.

  An example of the structure and corresponding signal flow of such a state of the art bandwidth extension system is illustrated in FIG. Usually, both the lower and upper frequency ranges are reintegrated.

  In block 601, the incoming or received acoustic signal x (n) in digitized form is processed by subsampling and block extraction to produce a signal vector.

を取得する。ここで、変数ｎは、時間を示す。この図において、入来信号ｘ（ｎ）は、サンプリングレートを増加させることによって、所望される帯域幅に既に変換されていることが仮定される。この変換ステップにおいて、生成されるべき追加の周波数コンポーネントは無く、このことは、例えば、適切なアンチエイリアシングフィルタエレメントまたはアンチイメージングフィルタエレメントを用いることによって、達成され得る。送信された信号を補正しないために、帯域幅拡張は、ミッシング周波数範囲内のみで行われる。送信方法に依存して、拡張は低周波数（例えば、０から３００Ｈｚまで）および／または高周波数（例えば、３４００Ｈｚから所望されるサンプリングレートの半分まで）の範囲に関係する。

To get. Here, the variable n indicates time. In this figure, it is assumed that the incoming signal x (n) has already been converted to the desired bandwidth by increasing the sampling rate. In this conversion step, there are no additional frequency components to be generated, which can be achieved, for example, by using suitable anti-aliasing or anti-imaging filter elements. In order not to correct the transmitted signal, the bandwidth extension is performed only within the missing frequency range. Depending on the transmission method, the extension relates to a range of low frequencies (eg, from 0 to 300 Hz) and / or high frequencies (eg, from 3400 Hz to half the desired sampling rate).

  At block 602, a narrowband spectral envelope is extracted from the narrowband signal, and the narrowband signal is limited by the bandwidth limitation of the telephone channel. A corresponding broadband envelope signal is estimated from the narrowband envelope via a non-linear mapping. The mapping is based on, for example, a set of code books (see Non-Patent Document 3) or a neutral network (see Non-Patent Document 4). In these methods, codebook entries or neutral network weights are generated using learning methods that require large processor and memory resources.

  Further, at block 603, a broadband or wideband excitation signal having a spectrally flat envelope is generated from the narrowband signal. This excitation signal corresponds to a signal recorded directly behind the vocal cords. That is, the excitation signal contains information about speech and pitch, but usually does not contain information about form and structure or spectral shape. Thus, in order to derive a complete signal (eg, a speech signal), the excitation signal needs to be weighted using a spectral envelope. Non-linear characteristics (see Non-Patent Document 5) (eg, two-beam rectification or squaring) can be used for generating the excitation signal.

  Excitation signal for bandwidth expansion

は、ブロック６０４において、包絡線を用いてスペクトル的に着色される。その後、拡張に対して使用されるスペクトル範囲は、ブロック６０６において帯域消去フィルタを用いて抽出され、信号ベクトル

Are spectrally colored using the envelope at block 604. The spectral range used for the extension is then extracted using a band elimination filter at block 606, and the signal vector

を生じる。帯域消去フィルタは、例えば、２００〜３７００Ｈｚまでの範囲において有効であり得る。

Produce. The band elimination filter may be effective in the range from 200 to 3700 Hz, for example.

  The signal vector of the received signal

は、ブロック６０５において、補足的な帯域通過フィルタを通過する。次いで、信号コンポーネント

Are passed through a supplemental bandpass filter at block 605. Then the signal component

が、追加され、拡張された帯域幅を有する信号ベクトル

Is added and a signal vector with expanded bandwidth

を取得する。ブロック６０７において、異なる信号ベクトルが、再び組み合わされ、オーバーサンプリングが行われ、信号ｙ（ｎ）を生じる。

To get. At block 607, the different signal vectors are recombined and oversampling is performed, resulting in signal y (n).

In these prior art systems, the elements and their parameters are not subsequently changed once implemented. Thus, all incoming acoustic signals are handled in the same way.
P. “Enhancement of Bandlimited Spech Signals” by Jax, 2002, Algorithms and Theoretical Bounds, Dissertation, Aachen, Germany E. Larsen, R.A. M.M. “Audio Bandwidth Extension” by Aarts, 2004, Hoboken, Wiley, New Jersey, USA J. et al. Epps, W.M. H. Holmes, “A New Technology for Wideband Enhancement of Coded Narrow-band Speech”, June 1999, IEEE Workshop on Speech Coding, 17th page, Con17 page. J. et al. M.M. Valin, R.A. Left Bandre, “Bandwidth Extension of Narrowband Speech for Low Bit-Rate Wideband Coding”, September 2000, IEEE Workshop on Speech Coding, Proceed 130 U. Kornagel, “Spectral Widening of the Excitation Signal for Telephone-Band Speech Enhancement”, September 2001, IWAENC01, Conference Proceedings, pages 215-218

  Accordingly, an object underlying the present invention is to provide a more flexible method and apparatus for providing an acoustic signal having an extended bandwidth.

  This problem is solved by the method according to item 1 and the device according to item 16.

In accordance with the present invention, a method for providing an acoustic signal having an extended bandwidth, the method comprising:
(A) automatically determining an upper limit and a lower limit of the current bandwidth of the received acoustic signal;
(B) in order to supplement the received acoustic signal, at least one supplemental signal between a lower limit of a predetermined wide bandwidth and the lower limit of the current bandwidth and / or of the current bandwidth; Automatically determining between an upper limit and an upper limit of a predetermined wideband bandwidth, wherein the lower limit of the predetermined wideband bandwidth is less than the lower limit of the current bandwidth, and the predetermined wideband bandwidth The bandwidth upper limit is greater than the current bandwidth upper limit,
(C) obtaining an acoustic signal having an extended bandwidth by automatically combining the at least one supplemental signal and the received acoustic signal.

  Determining the current bandwidth upper and lower bandwidth limits of the received acoustic signal, and between the current bandwidth limit and the respective predetermined wideband (or wideband) bandwidth limit By determining the supplementary signal at, the method according to the invention allows adaptation of the bandwidth extension to the actually received acoustic signal. For example, if the sender uses an ISDN phone, a wider frequency range is used compared to a mobile phone with a hands-free system. Therefore, the bandwidth of the received acoustic signal is expanded only at those frequencies so that the quality of the resulting signal is very high when it is needed.

  In this way, on the other hand, no spectral gap occurs even if the received signal covers only a very narrow frequency range. On the other hand, when receiving a signal that covers a relatively wide frequency range, the frequency is not cut off when determining the supplemental signal.

  The received acoustic signal is a digital signal or can be digitized. In the above method, steps (a)-(c) may be preceded by converting the received acoustic signal to a predetermined sampling rate. Furthermore, steps (a) to (c) may be preceded by a step of extracting a signal vector from the acoustic signal, in particular the transformed acoustic signal. The signal vector can be obtained by sub-sampling the acoustic signal and can comprise a predetermined number of entries. The subsequent signal vectors can then overlap. The use of signal vectors simplifies further processing of the signal.

  Steps (a)-(c) may be preceded by determining a spectral vector of the received acoustic signal. In particular, a window function can be applied to the signal vector of the received acoustic signal. For example, Hanning or Hamming windows can be used (see KD Kammeyer, K. Kroschel "Digital Signalverbeitung", 4th edition, 1997, Stuttgart, Tübner). The signal vector, in particular the signal vector thus weighted, can be transformed into the Fourier domain using a discrete Fourier transform. The resulting vector is a short-term spectral vector. This allows further processing in the Fourier domain.

  In the above method, step (b) includes determining the spectral envelope signal and the broadband excitation signal between a lower bandwidth limit and an upper bandwidth limit by determining the broadband spectral envelope signal and the broadband excitation signal. And the product of the excitation signal may correspond to an acoustic signal received according to a predetermined criterion.

  Such a decomposition into an envelope signal and an excitation signal simplifies the determination of the current bandwidth limit and increases the accuracy in determining the supplemental signal.

  Step (a) may include comparing the determined broadband spectral envelope signal to the long-term power spectrum of the received acoustic signal. It can be seen that the long-term power spectrum is a suitable basis for determining the current bandwidth limit of the acoustic signal.

  As a result, if the current bandwidth limit is thus determined in step (a) using the broadband spectral envelope signal of the received acoustic signal, the supplemental signal is determined in step (b). This is based on these current bandwidth limitations, and including the determination of the envelope signal is again by comparing the (newly) determined envelope signal with the long-term power spectrum. It is possible to adapt the current bandwidth limits repeatedly. In other words, determining the current bandwidth limit in step (a) uses the spectral envelope signal determined according to step (b), particularly in the preceding step or in the preceding iteration of the method. obtain.

  In particular, when the received acoustic signal is transformed into the Fourier domain, determining the long-term power spectrum is a first-order recursive smoothing of the square of the absolute value of the subband signal corresponding to the acoustic signal. It can include doing. This can be done in particular only if the desired signal (eg a speech signal) is detected in the received acoustic signal.

  Furthermore, the long-term power spectrum can be normalized, particularly with respect to the long-term power spectrum within predetermined frequency limits.

  Alternatively, the long-term power spectrum can be determined in the time domain. This can be done by determining autocorrelation and performing LPC analysis to obtain the corresponding prediction coefficients.

  The step of comparing is to select a minimum frequency and a maximum frequency, for which the long-term power spectrum is greater than the determined broadband spectrum envelope power spectrum plus a predetermined constant. Can be greater than or equal to.

  This is a particularly simple and reliable way to determine bandwidth limits. The predetermined constant can be selected based on experimental or theoretical data. The default constant can be a negative number.

  In the above method, determining a broadband spectral envelope signal may include selecting an envelope signal from a codebook according to predetermined criteria.

  By using a codebook, the desired computational power can be reduced to determine the envelope signal. In principle, different kinds of criteria can be used when selecting the envelope signal from the codebook. In particular, using a pre-determined distance criterion (eg, cepstrum distance) can be used, especially if the codebook entry has the form of a cepstrum vector.

  In particular, selecting the envelope signal equalizes the received acoustic signal and, in particular, has the smallest distance to the equalized acoustic signal according to a predetermined distance criterion, and in particular the smallest cepstrum distance. It may include selecting an envelope signal having from a code book.

  Equalizing the acoustic signal allows the acoustic signal to be modified so that the comparison with the envelope signal from the codebook can be simplified. In particular, the received acoustic signal may be equalized in such a way that the resulting signal exhibits a long-term power spectrum corresponding to the long-term power spectrum of the signal used for codebook learning. Equalization may be limited to frequencies between the upper limit of the current bandwidth of the received acoustic signal and the lower limit of the current bandwidth; outside these limits, the signal is not changed. In particular, the equalization of the received acoustic signal is performed using the normalized long-term power spectrum of the signal used for codebook learning (specifically, the normalized long-term power spectrum of the received acoustic signal itself). Can be performed using a normalized long-term power spectrum divided by the power spectrum.

  The codebook may comprise a plurality of sets of corresponding envelope signals, each set comprising a broadband envelope signal between a broadband bandwidth upper limit and a broadband bandwidth lower limit, A corresponding narrowband envelope signal is provided between the lower limit of the narrowband bandwidth greater than the lower limit of the width and the upper limit of the narrowband bandwidth smaller than the upper limit of the wideband bandwidth. Selecting a line signal determines a narrowband envelope signal that has a minimum distance to the equalized acoustic signal according to a predetermined distance criterion and determines the corresponding wideband envelope signal of this set. Selecting.

  In this way, a simple comparison between the received acoustic signal and the codebook elements can be made because the narrowband signal more closely matches the received acoustic signal, which usually has a narrowband bandwidth. Can be done.

  When using cepstrum distance to select an envelope signal, the received acoustic signal (especially its equalized form) needs to be converted into a cepstrum domain. As a result, the step of selecting the envelope signal is to determine the squared absolute value of the subband signal of the received acoustic signal, and in particular by performing an inverse discrete Fourier transform on the squared absolute value vector. , Further including determining autocorrelation in the time domain, determining a prediction coefficient, particularly using a Levinson-Durban algorithm, and performing recursion to obtain a cepstrum coefficient.

  In order to determine the spectral envelope from the cepstrum vector, the method recursively converts the cepstrum vector into prediction error coefficients, and adds a predetermined number of zeros to obtain the inverse spectrum, followed by It may further include increasing the prediction error filter vector by performing a discrete Fourier transform and determining the reciprocal of each subband component to obtain a spectral envelope vector.

  In the above method, the step of selecting an envelope signal is preceded by providing an adapted narrowband codebook envelope signal that is adapted to a lower limit of the current bandwidth and an upper limit of the current bandwidth. obtain.

  Such adaptation of the codebook entry allows an improved selection of the corresponding envelope signal from the codebook. In particular, if the received acoustic signal exhibits a wider bandwidth than the original narrowband envelope signal of the codebook, the adaptation results in an envelope signal having an extended bandwidth in the codebook. In this way, particularly frictional sounds can be detected more reliably.

  The providing step may include processing the wideband codebook envelope signal using the long-term power spectrum of the received acoustic signal.

  Due to the use of the power spectrum of the received acoustic signal, an appropriate fit to the acoustic signal can be obtained. The long-term power spectrum can be normalized; furthermore, the long-term power spectrum of the received acoustic signal can be divided by the normalized long-term power spectrum of the broadband signal used for codebook learning . The processing of the wideband codebook envelope signal can only be performed outside the current bandwidth limit; within the bandwidth limit, the envelope signal is not changed. The process using the long-term power spectrum may include weighting the wideband codebook envelope signal vector using the long-term power spectrum of the received acoustic signal.

  In the above method, determining the broadband excitation signal may be based on prediction error filtering and / or non-linear characteristics. In this way, an appropriate excitation signal can be generated. Possible nonlinear characteristics include, for example, U.S. Pat. It is disclosed in Kornagel, “Spectral Widening of the Excitation Signal for Telephone-Band Speech Enhancement”.

  In the above method, the at least one supplemental signal may be based on a product of the determined wideband spectral envelope and the determined wideband excitation signal, and step (c) may include a lower bound of the current bandwidth and the current Of the received acoustic signal between the upper bandwidth limit and the bandwidth between the lower bandwidth limit and the lower current bandwidth bandwidth, and / or the current upper bandwidth limit and the wide bandwidth bandwidth. It may include adding at least one supplemental signal limited to a band between the upper limit.

  As a result, the supplemental signal is based on spectrally coloring the excitation signal using the envelope signal. By adding the supplemental signal only outside the current bandwidth of the received acoustic signal, artifacts in the resulting signal with extended bandwidth are avoided.

  Step (c) may also include adapting the power of the supplemental signal and / or the received acoustic signal. In this step, the power of the received acoustic signal can be maintained.

  In the above method, at least one of the steps may be performed in a cepstrum region. In particular, if the codebook entry is a cepstrum vector, this allows the method to be performed in a simpler way.

  Steps (a) to (c) of the above method can be repeated at predetermined time intervals. The repeated adaptation to the currently received acoustic signal then leads to a permanently high quality of the resulting broadband signal.

  Steps (a) to (c) of the above method can be repeated only if the desired signal component (eg, speech behavior) is detected in the received acoustic signal. In particular in the case of speech signals, it is advantageous to expand the bandwidth of the received acoustic signal. As a result, limiting the method to detected speech behavior reduces the desired computational power and avoids the presence of artifacts due to imperfect fits.

  The present invention also provides a computer program product comprising one or more computer-readable media having computer-executable instructions for performing the above-described method steps when executed on a computer.

Further provided is an apparatus for providing an acoustic signal having an extended bandwidth, the apparatus comprising:
Bandwidth determining means for automatically determining an upper limit and a lower limit of the current bandwidth of the received acoustic signal;
The received acoustic signal between a lower limit of a predetermined wide bandwidth and the lower limit of the current bandwidth and / or between an upper limit of the current bandwidth and an upper limit of the predetermined wide bandwidth. To supplement, supplemental signal means for automatically determining at least one supplemental signal, the lower limit of the predetermined wideband bandwidth being less than the lower limit of the current bandwidth, A supplemental signal means, wherein the upper limit of the broadband bandwidth is greater than the upper limit of the current bandwidth;
The at least one supplemental signal and a combination means for acquiring an acoustic signal having an expanded bandwidth by automatically combining the received signal.

  Similar to the method described above, such a device provides an advantageous method for extending the bandwidth of the received acoustic signal. In particular, due to the determination of the current bandwidth upper limit and current bandwidth lower limit of the received acoustic signal and the determination of the corresponding supplemental signal, the quality of the resulting output signal is a band with fixed parameters. Increased compared to the width expansion system.

  The supplemental signal means determines the product of the spectral envelope signal and the excitation signal by determining the broadband spectral envelope signal and the broadband excitation signal between the upper bandwidth limit and the lower bandwidth limit. May be provided with means for responding to the received acoustic signal, according to predetermined criteria.

  The bandwidth determining means may be configured to compare the determined broadband spectral envelope signal with the long-term power spectrum of the received acoustic signal.

  The bandwidth determination means may be configured to select a minimum frequency and a maximum frequency, for which the long-term power spectrum adds a predetermined constant to the power spectrum of the determined wideband spectral envelope signal. Is greater than or equal to

  In the above apparatus, the means for determining the broadband spectral envelope signal may comprise means for selecting the envelope signal from the codebook according to predetermined criteria.

  The means for selecting the envelope signal has a minimum distance to the equalized acoustic signal, in particular a cepstrum distance, so as to equalize the received acoustic signal and according to a predetermined distance criterion An envelope signal may be configured to be selected from a code book.

  In the above apparatus, the codebook may comprise a plurality of sets of corresponding envelope signals, each set having a broadband envelope signal between a lower bandwidth limit and an upper bandwidth limit. A corresponding narrowband envelope between a lower limit of the narrowband bandwidth that is larger than the lower limit of the wideband bandwidth and an upper limit of the narrowband bandwidth that is smaller than the upper limit of the wideband bandwidth. And means for selecting an envelope signal to determine a narrowband envelope signal having a minimum distance relative to the equalized acoustic signal according to a predetermined distance criterion, and A set of corresponding wideband envelope signals may be selected.

  The means for determining the wideband spectral envelope signal comprises means for providing a adapted narrowband codebook envelope signal that is adapted to a lower limit of the current bandwidth and an upper limit of the current bandwidth. Can be prepared.

  The means for providing may be configured to process the wideband codebook envelope signal using the long-term power spectrum of the received acoustic signal.

  In the above apparatus, the means for determining the broadband excitation signal may be configured to determine the broadband excitation signal based on prediction error filtering and / or nonlinear characteristics.

  The at least one supplemental signal may be based on a product of the determined wideband spectral envelope and the determined excitation signal, and the combining means receives between the lower limit of the current bandwidth and the upper limit of the current bandwidth And / or at least one supplementary restriction between the lower limit of the wide bandwidth and the lower limit of the current bandwidth and / or between the upper limit of the current bandwidth and the upper limit of the wide bandwidth It may be configured to add the signal.

  In the above apparatus, at least one of the means may be configured to perform at least part of the function of the means in the cepstrum region.

  The means of the above apparatus may be configured to repeat each function of the means at predetermined time intervals.

  The apparatus may further comprise a desired signal detector, in particular a speech detector, so that the means perform the respective function of the means only if the desired signal component is detected in the received acoustic signal. Can be configured.

  The present invention further provides the following means.

(Item 1)
A method for providing an acoustic signal having an extended bandwidth, comprising:
(A) automatically determining an upper limit and a lower limit of the current bandwidth of the received acoustic signal;
(B) in order to supplement the received acoustic signal, at least one supplemental signal between a lower limit of a predetermined wide bandwidth and the lower limit of the current bandwidth and / or of the current bandwidth; Automatically determining between an upper limit and an upper limit of a predetermined wideband bandwidth, wherein the lower limit of the predetermined wideband bandwidth is less than the lower limit of the current bandwidth, and the predetermined wideband bandwidth The bandwidth upper limit is greater than the current bandwidth upper limit,
(C) obtaining an acoustic signal having an extended bandwidth by automatically combining the at least one supplemental signal and the received acoustic signal.

(Item 2)
The step (b) determines the spectral envelope signal and the excitation by determining the broadband spectral envelope signal and the broadband excitation signal between the lower limit of the broadband bandwidth and the upper limit of the broadband bandwidth. A method according to item 1, comprising a product with a signal corresponding to the received acoustic signal according to a predetermined criterion.

(Item 3)
3. The method of item 2, wherein step (a) comprises comparing the determined broadband spectral envelope signal to a long-term power spectrum of the received acoustic signal.

(Item 4)
The comparing step is to select a minimum frequency and a maximum frequency, for which the long-term power spectrum has a predetermined constant on the power spectrum of the determined wideband spectral envelope signal. 4. The method of item 3, comprising greater than or equal to the sum.

(Item 5)
5. A method according to one of items 2 to 4, wherein determining the broadband spectral envelope signal comprises selecting an envelope signal from a codebook according to predetermined criteria.

(Item 6)
Selecting the envelope signal equalizes the received acoustic signal and has a minimum distance to the equalized acoustic signal according to a predetermined distance criterion, in particular a minimum cepstrum distance. 6. The method of item 5, comprising selecting an envelope signal from the codebook.

(Item 7)
The codebook includes a plurality of sets of corresponding envelope signals, and each set includes a broadband envelope signal between the upper limit of the broadband bandwidth and the lower limit of the broadband bandwidth. A corresponding narrowband envelope signal between a lower limit of the narrowband bandwidth that is larger than the lower limit of the wideband bandwidth and an upper limit of the narrowband bandwidth that is smaller than the upper limit of the wideband bandwidth With
Selecting an envelope signal includes determining a narrowband envelope signal having a minimum distance relative to the equalized acoustic signal according to the predetermined distance criterion, and the corresponding wideband signal of the set. 7. The method of item 6, comprising selecting an envelope signal.

(Item 8)
The step of selecting the envelope signal is preceded by providing an adapted narrowband codebook envelope signal that is adapted to the lower limit of the current bandwidth and the upper limit of the current bandwidth, 8. The method according to item 7.

(Item 9)
9. The method of item 8, wherein the providing step comprises processing a wideband codebook envelope signal using a long-term power spectrum of the received acoustic signal.

(Item 10)
10. The method according to one of items 2 to 9, wherein determining the broadband excitation signal is based on prediction error filtering and / or nonlinear characteristics.

(Item 11)
The at least one supplemental signal is based on a product of the determined broadband spectral envelope and the determined broadband excitation signal, and the step (c) includes the lower limit of the current bandwidth and the current The received acoustic signal between the upper bandwidth limit and the bandwidth between the lower bandwidth limit and the lower current bandwidth bandwidth and / or the current upper bandwidth bandwidth and the upper bandwidth bandwidth 11. A method according to one of items 2 to 10, comprising summing the at least one supplemental signal limited to a band between a broadband bandwidth upper limit.

(Item 12)
Item 12. The method of item 1 to item 11, wherein at least one of the steps is performed in a cepstrum region.

(Item 13)
13. The method according to one of items 1 to 12, wherein the steps (a) to (c) are repeated at predetermined time intervals.

(Item 14)
Steps (a) to (c) are repeated as described in one of items 1 to 13, only when a desired signal component, in particular speech behavior, is detected in the received acoustic signal. the method of.

(Item 15)
One or more computer-readable media having computer-executable instructions for performing the above steps of the method of claim 1 when executed on a computer. , Computer program products.

(Item 16)
An apparatus for providing an acoustic signal having an extended bandwidth, comprising:
Bandwidth determining means for automatically determining an upper limit and a lower limit of the current bandwidth of the received acoustic signal;
The received acoustic signal between a lower limit of a predetermined wide bandwidth and the lower limit of the current bandwidth and / or between an upper limit of the current bandwidth and an upper limit of the predetermined wide bandwidth. To supplement, supplemental signal means for automatically determining at least one supplemental signal, the lower limit of the predetermined wideband bandwidth being less than the lower limit of the current bandwidth, A supplemental signal means, wherein the upper limit of the broadband bandwidth is greater than the upper limit of the current bandwidth;
An apparatus comprising: combining means for obtaining an acoustic signal having an expanded bandwidth by automatically combining the at least one supplemental signal and the received signal.

(Item 17)
The supplemental signal means determines the spectral envelope signal and the excitation signal by determining a broadband spectral envelope signal and a broadband excitation signal between a lower limit of the broadband bandwidth and an upper limit of the broadband bandwidth. Item 17. The device of item 16, wherein the product with the signal comprises means corresponding to the received acoustic signal according to a predetermined criterion.

(Item 18)
Item 18. The apparatus of item 17, wherein the bandwidth determining means is configured to compare the determined broadband spectral envelope signal with a long-term power spectrum of the received acoustic signal.

(Item 19)
The bandwidth determining means is configured to select a minimum frequency and a maximum frequency, and for the frequency, the long-term power spectrum is predetermined as a power spectrum of the determined broadband spectrum envelope signal. Item 19. The device of item 18, wherein the device is greater than or equal to a constant of

(Item 20)
Item 20. The item of items 17-19, wherein the means for determining a broadband spectral envelope signal comprises means for selecting an envelope signal from a codebook according to predetermined criteria. apparatus.

(Item 21)
The means for selecting an envelope signal has a minimum distance relative to the equalized acoustic signal so as to equalize the received acoustic signal and according to a predetermined distance criterion, in particular a minimum 21. The apparatus of item 20, wherein the apparatus is configured to select an envelope signal having a cepstrum distance of from the codebook.

(Item 22)
The codebook includes a plurality of sets of corresponding envelope signals, and each set includes a broadband envelope signal between the lower limit of the broadband bandwidth and the upper limit of the broadband bandwidth. A corresponding narrowband envelope signal between a lower limit of the narrowband bandwidth greater than the lower limit of the wideband bandwidth and an upper limit of the narrowband bandwidth smaller than the upper limit of the wideband bandwidth With
The means for selecting an envelope signal is adapted to determine a narrowband envelope signal having a minimum distance relative to the equalized acoustic signal according to the predetermined distance criterion, and for the set of the above Item 22. The device of item 21, wherein the device is configured to select a corresponding broadband envelope signal.

(Item 23)
The means for determining the wideband spectral envelope signal provides a adapted narrowband codebook envelope signal that is adapted to the lower limit of the current bandwidth and the upper limit of the current bandwidth. Item 23. The apparatus according to Item 22, comprising the following means.

(Item 24)
24. The apparatus of item 23, wherein the means for providing is configured to process the broadband codebook envelope signal using a long-term power spectrum of the received acoustic signal.

(Item 25)
Item 1 to Item 24 wherein the means for determining a broadband excitation signal is configured to determine the broadband excitation signal based on prediction error filtering and / or nonlinear characteristics. The device described.

(Item 26)
The at least one supplemental signal is based on a product of the determined broadband spectral envelope and the determined broadband excitation signal, and the combining means includes a lower limit of the current bandwidth and the current bandwidth. The received acoustic signal between the upper limit of the width, the band between the lower limit of the wide bandwidth and the lower limit of the current bandwidth, and / or the upper limit of the current bandwidth and the wide band 26. The apparatus according to one of items 17 to 25, wherein the apparatus is configured to add the at least one supplemental signal limited to a band between an upper bandwidth limit.

(Item 27)
27. Apparatus according to one of items 16 to 26, wherein at least one of said means is arranged to perform at least part of the function of said means in said cepstrum region.

(Item 28)
28. The apparatus according to one of items 16 to 27, wherein the means is configured to repeatedly perform each function of the means at a predetermined time interval.

(Item 29)
Further comprising a desired signal detector, in particular a speech detector, said means performing the respective function of said means only if the desired signal component is detected in said received acoustic signal. 29. The apparatus according to one of items 16 to 28, wherein the device is configured as follows.

(Summary)
The present invention is directed to a method for providing an acoustic signal having an extended bandwidth, the method automatically automating a current bandwidth upper limit and a current bandwidth lower limit of a received acoustic signal. In order to determine and supplement the received acoustic signal at least one supplemental signal between a predetermined lower bandwidth limit and a lower current bandwidth limit and / or a current bandwidth The lower limit of the predetermined wideband bandwidth is smaller than the lower limit of the current bandwidth, and the predetermined upper limit of the predetermined wideband bandwidth The broadband bandwidth upper limit is greater than the current bandwidth upper limit, and automatically combining the at least one supplemental signal and the received acoustic signal to increase the expanded bandwidth. Obtaining an acoustic signal having .

  Further features and advantages of the invention are described below with reference to the drawings.

  FIG. 1 shows the structure of signal flow in an apparatus for providing an acoustic signal having an extended bandwidth. FIG. 2 is a flow diagram illustrating an example of a method for providing an acoustic signal having an extended bandwidth, which may be performed by an apparatus corresponding to FIG. In view of this, FIGS. 1 and 2 are simultaneously described below.

  According to step 201, an acoustic signal (eg, a speech signal) is received over a telephone line. Since telephone line bandwidth is limited, bandwidth expansion is desired to improve signal quality. Therefore, the signal should be increased to obtain a predefined wider bandwidth. It will be appreciated that the methods described below can be used for bandwidth expansion independent of the type of input signal and independent of the type of transmission line (ie, no telephone line need be present). Should be.

  The acoustic signal x (n) received by block 101 has already been pre-processed by increasing the sampling rate to a predetermined wideband or narrowband bandwidth. However, in this way no additional frequency component is generated. This can be accomplished, for example, using a suitable anti-aliasing filter or anti-imaging filter. This type of bandwidth expansion is preferably done only for the “missing” frequency region; in the case of analog telephone lines, these regions can be between 0 Hz and 300 Hz and are desired from 3400 Hz to It can be up to half the sampling rate (eg, up to 3700 Hz).

  From the resulting signal x (n), where n represents a time variable, a signal vector

が生成される（ステップ２０２）。このことは、特定の長さになるまでｒごとにサンプリング値をとることによって、達成され得る。従って、Ｎ_ａｎａ個のエレメントを有する信号ベクトルは以下：

Is generated (step 202). This can be achieved by taking a sampling value every r until a specific length is reached. Thus, a signal vector with _Nana elements is:

の形式を有する。

Has the form

It should be noted that there can be overlap between adjacent signal vectors. For a desired or final sampling rate of 11.025 kHz, the following values:
r = 64,
_Nana = 256
Can be taken.

  Thereafter (step 203), a windowing procedure is performed on the signal vector so that a windowed signal vector is obtained.

が取得される。

Is acquired.

  Window matrix

は、

Is

の形式の対角行列である。

Is a diagonal matrix of the form

  The elements of this matrix can be selected to correspond to different types of windows. A typical window is a Hanning or Hamming window. The weighted signal vector is a discrete Fourier transform:

を用いて、フーリエ領域に変換される。

Is converted into the Fourier domain.

  The resulting short-term spectral vector has the form:

を有し、ここで、Ω_μは、周波数変数を示す。

Where Ω _μ denotes a frequency variable.

  Based on the spectral vector, a long-term power spectrum of the received acoustic signal is determined at block 102 (step 204). There are different possibilities for estimating such a long-term power spectrum. According to one alternative, first order recursive smoothing is applied to the subband signal.

の絶対値の二乗において行われる。

This is done in the square of the absolute value of.

好ましくは、時間定数β_ｆｒｅは、１に近い（０≪β_ｆｒｅ＜１）ように選択されることにより、十分に大きい平均時間が取得される。

Preferably, the time constant β _fre is selected to be close to 1 (0 _<< β _fre <1), thereby _obtaining a sufficiently large average time.

  In principle, the recursive smoothing according to the first line of the above equation can be performed continuously. However, to avoid any artifacts, recursive smoothing can be performed only if the desired signal component is present in the received acoustic signal (eg, when speech behavior is detected). For this purpose, a speech detector is for example described in E.I. Haensler, G.M. As described by Schmidt, “Acoustic Echo and Noise Control-A Practical Approach”, Hoboken, NJ, 2004, Wiley.

  To simplify further processing, the long-term power spectrum can be normalized to long-term power within a predetermined frequency band.

帯域限界

Bandwidth limit

は、所定の周波数帯域の下限および上限を示す。例えば、この周波数帯域は、本方法が使用されるべきである最小の帯域幅を有する電話帯域に対応し得る（例えば、限界は４００Ｈｚおよび３３００Ｈｚであり得る）。好ましくは、限界は帯域に対応し、帯域は狭い周波数帯域の周波数帯域より小さいかまたはほぼ等しく、該狭い周波数帯域内で、以下に記載されるコードブックが、学習されている。これらの限界はΩ_ｌおよびΩ_ｕによって示されている。

Indicates the lower and upper limits of the predetermined frequency band. For example, this frequency band may correspond to the telephone band having the smallest bandwidth that the method should be used (eg, the limits may be 400 Hz and 3300 Hz). Preferably, the limit corresponds to a band, which is less than or approximately equal to the frequency band of the narrow frequency band, within which the codebook described below is learned. These limits are indicated by Ω _l and Ω _u .

  Alternatively, estimation can be done in the time domain as well to determine a long-term power spectrum in the frequency domain. For this purpose, autocorrelation is estimated for approximately 10-20 sampling cycles of offset. The prediction coefficients can then be determined using LPC (Linear Predictive Coding) analysis. The long-term power spectrum is obtained via a discrete Fourier function and division.

  In block 103 (step 205), the acoustic signal is equalized. Equalization:

は、上で決定されたスペクトルベクトルに対して行われる。

Is performed on the spectral vector determined above.

  Equalization matrix

は形式

Is the format

の対角行列であり、エントリ

A diagonal matrix of entries

を有する。上記の数式において、

Have In the above formula,

は、受信された音響信号の現在の帯域幅の下限および上限を示す。従って、アップデートされ等化された信号を取得するために、時間（ｎ−１）における帯域幅の限界は、現在の帯域幅の限界をとる。さらに、

Indicates the lower and upper limits of the current bandwidth of the received acoustic signal. Thus, to obtain an updated and equalized signal, the bandwidth limit at time (n−1) takes the current bandwidth limit. further,

は、広帯域の信号の正規化された長期のパワースペクトルを示し、これは、コードブックの学習のために使用されている。このようなパワースペクトルの正規化は、上記の受信された音響信号の長期のパワースペクトルの場合に類似して行われる。コードブックを学習するために使用されるこのような正規化された長期のパワースペクトルに対する例が、図３に示される。

Indicates the normalized long-term power spectrum of a wideband signal, which is used for codebook learning. Such normalization of the power spectrum is performed similarly to the case of the long-term power spectrum of the received acoustic signal. An example for such a normalized long-term power spectrum used to learn a codebook is shown in FIG.

Equalization is a minimum and maximum value, for example
H _{eq, min} = -12 dB
H _{eq, max} = 12 dB
Limited to

  As can be seen from the above, the acoustic signal is equalized at the previous time step only within the limits of the current bandwidth. No equalization is performed outside these bandwidths.

  In the following, determining the broadband spectral envelope will be described in more detail. An envelope signal corresponding to the received acoustic signal is determined using a code book. The codebook used has many pairs of corresponding narrowband and wideband envelope signals. The codebook was acquired by learning using a large database based on the start of a long-term power spectrum (Y. Linde, A. Buzo, RM Gray “An Algorithm for Vector Quantizer Design”, January 1980, see IEEE Trans. Comm., COM-28, No. 1, pages 84-95).

  As shown in FIG. 2, the codebook entry is adapted at step 206 (block 104). In particular, narrowband codebook entries

が適合される。

Is adapted.

  This is accomplished by initiating a broadband entry in the codebook. The broadband envelope signal is the cepstrum vector

として提供される場合には、対応するスペクトル

If provided as the corresponding spectrum

が決定される。これらの広帯域のスペクトル包絡線に基づいて、適合されたまたは最適化された狭帯域のスペクトルが、重みづけ行列と掛けることによって決定される。

Is determined. Based on these broadband spectral envelopes, an adapted or optimized narrowband spectrum is determined by multiplying with a weighting matrix.

重みづけ行列は、形式：

The weighting matrix has the form:

で、エントリ

And the entry

を有する対角行列である。

Is a diagonal matrix.

  The cepstrum vector is then determined from the resulting spectral narrowband envelope.

  The conversion from a spectral vector to a cepstrum vector and vice versa is described below with respect to step 207, where a broadband spectral envelope is determined (block 105).

  The broadband spectral envelope from the codebook that best matches the acoustic signal is determined by comparing the narrowband codebook entry (after equalization) with the spectral envelope of the acoustic signal. The narrowband codebook entry is selected to have a small distance to the acoustic signal spectrum. In principle, different distance criteria can be used. Cepstrum distance is particularly useful because codebook entries are provided in the form of cepstrum vectors.

  When the optimal narrowband codebook entry is selected, the corresponding wideband codebook entry is determined as the optimal wideband spectral envelope for the received acoustic signal. As mentioned above, due to the adaptation of the wideband codebook entry, the optimal narrowband envelope can be selected in a very reliable manner.

In particular, converting the spectral vector of the received acoustic signal into a cepstrum vector is accomplished by:
1. Each subband signal

の二乗された絶対値を決定すること。
２．このベクトルに逆離散フーリエ変換を適用することであって、時間領域における自己相関の推定を生じること。
３．レビンソン−ダーバンアルゴリズムを用いて、予測係数（約１０〜２０の次数を有する）を、自己相関から決定すること。
４．次数に関する帰納を行うことによって、予測係数は、ケプストラム係数を決定するために使用される。通常、次数は予測される次数の１．５倍に対応する。

To determine the squared absolute value of.
2. Applying an inverse discrete Fourier transform to this vector, resulting in an estimation of autocorrelation in the time domain.
3. Determining prediction coefficients (having an order of about 10-20) from autocorrelation using the Levinson-Durban algorithm.
4). By performing induction on the order, the prediction coefficients are used to determine the cepstrum coefficients. Usually, the order corresponds to 1.5 times the expected order.

  The optimal cepstrum vector for a wideband codebook is

によって指定される。結果として生じるスペクトル包絡線は、形式：

Specified by. The resulting spectral envelope has the form:

を有する。

Have

Conversion of the cepstrum vector to a spectral vector is accomplished by:
1. Obtain the prediction error filter coefficients by transforming the cepstrum vector (as described above) using induction on the order.
2. Obtaining the inverse spectrum by incrementing the prediction error filter vector by a predetermined number of zeros followed by a discrete Fourier function.
3. A vector by determining the reciprocal of each subband component

を生成すること。０による除法は、例えば、適切な定数を追加することによって、個別に扱われる必要がある。

To generate. The division by zero needs to be handled individually, for example by adding appropriate constants.

  FIG. 4 illustrates an example of a codebook having four sets of entries. In each figure, the corresponding original narrowband envelope and the corresponding adapted narrowband envelope are shown. The original broadband and narrowband codebook entries are obtained based on a large database for ISDN telephone connections. As can be seen in this figure, the resulting optimized entry after fitting has a higher upper frequency limit. This allows for improved frictional sound detection.

  In step 208 (block 103), an excitation signal corresponding to the received acoustic signal is generated. This broadband excitation signal exhibits a spectrally flat envelope. It corresponds to a signal recorded directly behind the vocal cords.

  To determine the broadband excitation signal, first the equalized short-term spectrum

のスペクトル包絡線が、予測誤差フィルタ係数の形式で推定される。逆離散フーリエ変換をこのスペクトルベクトルに適用することは、対応する時間信号を決定することを可能にする。その後、時間領域におけるベクトルは、予測誤差フィルタによってフィルタされる。対応するフィルタ係数は、先に決定されているフィルタ係数である。

Are estimated in the form of prediction error filter coefficients. Applying an inverse discrete Fourier transform to this spectral vector makes it possible to determine the corresponding time signal. The vectors in the time domain are then filtered by a prediction error filter. The corresponding filter coefficient is a previously determined filter coefficient.

  Non-linear characteristics (eg, bi-directional rectification or squaring) are then applied to the filtered time domain vector. This creates missing low frequency and high frequency signal components. The transform in the Fourier domain is the expanded spectrum of the excitation signal

を提供する。

I will provide a.

  Alternatively, determining the excitation signal can be done in the time subband or Fourier domain as well. Examples for this alternative are Iser, G.M. Schmidt “Bandwidth Extension of Telephony Speech”, June 2005, Europe Newsletter, Volume 16, Number 2, pages 2-24.

  In the following step 209 (block 107), the broadband spectral envelope and the excitation signal are used to spectrally color the excitation signal. This can be accomplished by multiplication in the subband or Fourier domain:

対角行列

Diagonal matrix

は、形式：

The format:

を有する。

Have

  When generating the excitation signal, the power of the acoustic signal need not be maintained due to non-linearity or prediction error filtering. Therefore, power adaptation can be performed:

補正率Ｋは、以下のように選択され得、

The correction factor K can be selected as follows:

ここで

here

は、上記の長期のパワースペクトルの推定における場合と同一の帯域幅の限界を示す。

Indicates the same bandwidth limit as in the long-term power spectrum estimation described above.

  The current bandwidth limit is met at step 210 (block 108). According to one possibility, the bandwidth limit is determined at the start of a comparison between the spectrum of the received acoustic signal and a broadband spectral envelope that is subtracted by a predetermined constant:

パラメータＫ_Ｃは、値：
Ｋ_Ｃ＝−１２ｄＢ
を有し得る。

Parameter _{K C} is, the value:
K _C = -12 dB
Can have.

  In FIG. 5, an example for determining bandwidth limits is illustrated. The above intermediate limit is given by the intersection between the lowered broadband spectral envelope and the spectrum of the received acoustic signal.

  These intermediate limits can be smoothed recursively to remove incomplete temporal estimates. In this case, preferably smoothing is performed only if speech behavior is detected in the current signal frame.

次いで、受信された音響信号は、適応帯域通過フィルタを通過し、現在の帯域幅の限界内のコンポーネントのみを保持し（ブロック１０９）、スペクトルベクトル

The received acoustic signal is then passed through an adaptive bandpass filter, retaining only the components within the current bandwidth limit (block 109), and the spectral vector.

が取得される。同様に、スペクトル的に着色された励起信号は、補足的な適応帯域消去フィルタを通過する（ブロック１１０）ことにより、ベクトル

Is acquired. Similarly, the spectrally colored excitation signal is passed through a supplemental adaptive band elimination filter (block 110) to produce a vector.

が取得される。

Is acquired.

  By starting the addition of these two spectral vectors, an output signal having a standard bandwidth is generated (step 211):

これらのベクトルのコンポーネントは：

These vector components are:

として生成され、ここで、重みづけ行列

Where is the weighting matrix

は、対角行列：

Is a diagonal matrix:

である。

It is.

  matrix

の要素は：

The elements of are:

として決定され、補足重みづけ行列の重みは、加算された際に単位行列を生じるように決定される。

And the weights of the supplemental weighting matrix are determined to produce a unit matrix when added.

あるいは、帯域幅限界における遷移が、より平滑な方法で実現され得る。

Alternatively, the transition at the bandwidth limit can be realized in a smoother manner.

  Resulting output spectrum

は、次いで、逆フーリエ変換：

Then the inverse Fourier transform:

を介して時間領域に変換され、結果のベクトルをウィンドウイングすることが続く。特に、Ｎ_ａｎａおよびｒに対する上記された値およびハニングウィンドウを用いる場合には、このウィンドウ関数が、ウィンドウイングされた時間領域ベクトル：

Followed by windowing the resulting vector. In particular, when using the values and Hanning windows described above for _Nana and r, this window function is a windowed time domain vector:

を取得するために再び使用され得る。

Can be used again to get

  The resulting time domain vectors are then combined using an overlap addition method (as described in KD Kammeyer, K. Kroschel "Digital Signalverbeitung").

  In the above steps of the method, a more complex filter bank system can be used instead of the conventional discrete Fourier transform and inverse discrete Fourier transform (eg, PP Vaidyanathan “Multirate Systems and Filter Banks”, Parent Hall). 1992, Englewood Cliffs, New Jersey, USA).

  Further alternatives to the above variants are possible as well. For example, steps performed in the Fourier domain can also be performed in the time domain. Furthermore, equalizing the acoustic signal can be done in adapting a narrowband codebook entry. Furthermore, the above equalization step can be enhanced. For example, if amplification or attenuation is detected at a particular frequency, it can be adjusted within bandwidth limits. In this case, the output vector

は、重みづけ行列

Is the weighting matrix

を用いて変更される。

It is changed using.

  In addition to the above codebook analysis for broadband spectral envelopes, so-called linear mappings (see B. Iser, G. Schmidt, “Bandwidth Extension of Telephony Speech”) can additionally be used.

  Further modifications and variations of the present invention will be apparent to those skilled in the art upon review of this description. Accordingly, this description is to be construed as illustrative only and is for the purpose of teaching those skilled in the art the general manner of carrying out the invention. It should be understood that the form of the invention shown and described herein is to be taken as the presently preferred embodiment.

FIG. 1 illustrates the structure of an example apparatus for providing an acoustic signal having an extended bandwidth. FIG. 2 is a flow diagram of an example method for providing an acoustic signal having an extended bandwidth. FIG. 3 illustrates an example of a normalized long-term power spectrum for learning a codebook. FIG. 4 illustrates an example of a codebook entry. FIG. 5 illustrates the determination of the current bandwidth limit. FIG. 6 illustrates the structure of a prior art system.

Claims (23)

A method for providing an acoustic signal having an extended bandwidth, comprising:
(A) automatically determining an upper limit and a lower limit of the current bandwidth of the received acoustic signal;
(B) in order to supplement the received acoustic signal, at least one supplemental signal between a lower limit of a predetermined wide bandwidth and the lower limit of the current bandwidth and / or of the current bandwidth; Automatically determining between an upper limit and an upper limit of a predetermined wideband bandwidth, wherein the lower limit of the predetermined wideband bandwidth is less than the lower limit of the current bandwidth, and the predetermined wideband bandwidth The bandwidth upper limit is greater than the current bandwidth upper limit,
(C) obtaining an acoustic signal having an extended bandwidth by automatically combining the at least one supplemental signal and the received acoustic signal ;
The step (b) determines the spectral envelope signal and the excitation by determining the broadband spectral envelope signal and the broadband excitation signal between a lower limit of the broadband bandwidth and an upper limit of the broadband bandwidth. Including a product with a signal corresponding to the received acoustic signal according to predetermined criteria,
The step (a) comprises comparing the determined broadband spectral envelope signal with the spectrum of the received acoustic signal to determine the upper limit of the current bandwidth and the lower limit of the current bandwidth. Including repeatedly adapting,
The comparing step is to select a minimum frequency and a maximum frequency, for which the spectrum is greater than the determined wideband spectral envelope signal plus a predetermined constant , Or equal to . The method of claim 1 , wherein determining the wideband spectral envelope signal comprises selecting an envelope signal from a codebook according to predetermined criteria. Selecting the envelope signal equalizes the received acoustic signal and has a minimum distance to the equalized acoustic signal according to a predetermined distance criterion, in particular a minimum cepstrum distance. 3. The method of claim 2 , comprising selecting an envelope signal from the codebook. The codebook includes a plurality of sets of corresponding envelope signals, and each set includes a broadband envelope signal between an upper limit of the broadband bandwidth and a lower limit of the broadband bandwidth. A corresponding narrowband envelope signal between a lower limit of the narrowband bandwidth that is larger than the lower limit of the wideband bandwidth and an upper limit of the narrowband bandwidth that is smaller than the upper limit of the wideband bandwidth With
Selecting an envelope signal includes determining a narrowband envelope signal having a minimum distance relative to the equalized acoustic signal according to the predetermined distance criterion, and the corresponding wideband signal of the set. 4. The method of claim 3 , comprising selecting an envelope signal. Selecting the envelope signal is preceded by providing an adapted narrowband codebook envelope signal that is adapted to the lower limit of the current bandwidth and the upper limit of the current bandwidth; The method of claim 4 . 6. The method of claim 5 , wherein the providing step includes processing a wideband codebook envelope signal using a long-term power spectrum of the received acoustic signal. It is based on prediction error filtering and / or non-linear characteristics, the method according to one of the claims 1 to 6 for determining a broadband excitation signal. The at least one supplemental signal is based on a product of the determined broadband spectral envelope and the determined broadband excitation signal, and the step (c) includes the lower limit of the current bandwidth and the current The received acoustic signal between the upper bandwidth limit and the bandwidth between the lower bandwidth limit and the lower bandwidth limit and / or the current bandwidth upper limit and said at least limited to a band between the upper limit of the bandwidth of the wideband involves adding the single supplemental signal method as recited in one of claims 1 to 7. The method according to one of claims 1 to 8 , wherein at least one of the steps is performed in a cepstrum region. Wherein step (a) ~ (c) are repeated at predetermined time intervals, The method as recited in one of claims 1 to 9. Wherein step (a) ~ (c) is desired signal component, in particular a speech act, the repeated only if it is detected in the received acoustic signal, one of Claim 1 to claim 10 The method described in 1. When running on a computer, according to claim 1 wherein the computer-readable recording medium body computer-executable instructions for performing the steps of the method according to one of the claims 11. An apparatus for providing an acoustic signal having an extended bandwidth, comprising:
Bandwidth determining means for automatically determining an upper limit and a lower limit of the current bandwidth of the received acoustic signal;
The received acoustic signal between a lower limit of a predetermined wide bandwidth and the lower limit of the current bandwidth and / or between an upper limit of the current bandwidth and an upper limit of the predetermined wide bandwidth. To supplement, supplemental signal means for automatically determining at least one supplemental signal, the lower limit of the predetermined wideband bandwidth being less than the lower limit of the current bandwidth, A supplemental signal means, wherein the upper limit of the broadband bandwidth is greater than the upper limit of the current bandwidth;
Combining means for obtaining an acoustic signal having an expanded bandwidth by automatically combining the at least one supplemental signal and the received signal ;
The supplementary signal means determines the spectral envelope signal and the excitation signal by determining a broadband spectral envelope signal and a broadband excitation signal between a lower limit of the broadband bandwidth and an upper limit of the broadband bandwidth. Means for causing a product with the signal to correspond to the received acoustic signal according to a predetermined criterion;
The bandwidth determining means compares the determined broadband spectral envelope signal with the spectrum of the received acoustic signal, thereby determining the upper limit of the current bandwidth and the lower limit of the current bandwidth. Is configured to repeatedly adapt,
The bandwidth determining means is configured to select a minimum frequency and a maximum frequency, for which the spectrum is the determined broadband spectral envelope signal plus a predetermined constant. Greater than or equal to the device. 14. The apparatus of claim 13 , wherein the means for determining a broadband spectral envelope signal comprises means for selecting an envelope signal from a codebook according to predetermined criteria. The means for selecting an envelope signal has a minimum distance with respect to the equalized acoustic signal so as to equalize the received acoustic signal and according to a predetermined distance criterion, in particular a minimum 15. The apparatus of claim 14 , wherein the apparatus is configured to select an envelope signal having a cepstrum distance of from the codebook. The codebook includes a plurality of sets of corresponding envelope signals, and each set includes a broadband envelope signal between the lower limit of the broadband bandwidth and the upper limit of the broadband bandwidth. A corresponding narrowband envelope signal between a lower limit of the narrowband bandwidth that is larger than the lower limit of the wideband bandwidth and an upper limit of the narrowband bandwidth that is smaller than the upper limit of the wideband bandwidth With
The means for selecting an envelope signal is adapted to determine a narrowband envelope signal having a minimum distance to the equalized acoustic signal according to the predetermined distance criterion, and the set of the The apparatus of claim 15 , configured to select a corresponding broadband envelope signal. The means for determining the wideband spectral envelope signal provides a adapted narrowband codebook envelope signal that is adapted to the lower limit of the current bandwidth and the upper limit of the current bandwidth. The apparatus according to claim 16 , comprising: The apparatus of claim 17 , wherein the means for providing is configured to process the wideband codebook envelope signal using a long-term power spectrum of the received acoustic signal. Said means for determining a broadband excitation signal is configured to determine an excitation signal of the wideband based on prediction error filtering and / or non-linear characteristics, one of claim 13 to claim 18 The device according to item. The at least one supplemental signal is based on a product of the determined broadband spectral envelope and the determined broadband excitation signal, and the combining means includes: a lower limit of the current bandwidth; and the current bandwidth The received acoustic signal between the upper bandwidth limit, the bandwidth between the lower bandwidth limit and the lower current bandwidth bandwidth, and / or the current upper bandwidth bandwidth and the wide bandwidth bandwidth. The apparatus according to one of claims 13 to 19 , configured to add the at least one supplemental signal limited to a band between an upper limit of bandwidth. At least one, at least a part of the function of said means is configured to perform in the cepstral domain, according to one of claims 13 to claim 20 of said means. The apparatus according to any one of claims 13 to 21 , wherein the means is configured to repeat each function of the means at predetermined time intervals. Further comprising a desired signal detector, in particular a speech detector, said means performing the respective function of said means only if a desired signal component is detected in said received acoustic signal. The apparatus according to any one of claims 13 to 22 , which is configured as follows.

Images