EP1130577A2 - Method for the reconstruction of low speech frequencies from mid-range frequencies - Google Patents

Abstract

The invention relates to a method and a device for the reconstruction of low-frequency speech components from medium-high frequency components, in which the technical problem of enabling the reproduction of a disrupted speech signal to be as close to nature as possible is solved by a method in which above a cut-off frequency ω _0, at least two Adjacent frequency components ω _fa1 , ω _fa2 , ... are determined with increased amplitude in the speech signal and in which the fundamental frequency ω _{g of} the speech signal is determined as the frequency difference between the at least two adjacent frequency components ω _fa1 , ω _fa2 , ... and in which the low-frequency range below the cut-off frequency ω _{g is} reconstructed with the aid of the determined fundamental frequency ω _g and the speech signal.

The invention also relates to a device for performing the aforementioned method.

Description

The invention relates to a method and a device for reconstruction low-frequency speech components from medium-high frequency components.

In the prior art of digital processing of voice signals with a high Noise level in the low frequency range, the signal is improved by either Interference components are filtered out or very strongly disturbed frequency range from the Signal can be completely filtered out.

From US 5,842,160 A is a method for improving the quality of a digital Voice transmission known, depending on the different frequency bands Energy content can be assigned to different amounts of data. By the nature of the Coding and transmission result in low-energy signal areas that too Lead gaps in the received signal spectrum. These gaps are covered by Signals filled synthetically from the available data, so that a natural sounding speech signal is achieved.

From US 4,091,237 A is a method for determining the basic voice frequency of a digital voice signal known in real time. Especially for signals with one restricted frequency range, such as phone signals, and with a high The speech signal is improved by reducing noise by filtering out noise become. The signal is split up and in by a plurality of bandpass filters corresponding histogram is formed from which the basic voice frequency is extracted. If the fundamental frequency is known, noise can be recognized by the fact that it is in have no harmonic relationship to the fundamental frequency. The previously described The method is used to determine the basic frequency characteristic of a voice determine.

Furthermore, DE 37 33 983 describes a method for damping interference signals in a hearing aid known in which the signal is digitized and in individual Frequency ranges is divided. Frequency ranges with certain characteristics, such as rapid or very slow changes in the spectral distribution are damped and / or the cutoff frequencies are shifted. The signal cleaned in this way is in converted synthetic speech signals.

The method described above and the associated devices is the The disadvantage is that the speech signal is not at all or only in an insufficient form is reconstructed in order to generate the most natural possible output speech signal.

The methods described above can be used, among other things, for digital Voice amplification (digital voice enhancement - DVE) can be used. For example, there are two microphones above each row of seats in a motor vehicle attached so that it is possible. All vehicle occupants, for example, on one Participate in phone call. The system transmits the language to the front the microphone was recorded on the rear series speakers and vice versa. The system is therefore full with the handsfree telephone and the Radio / CD / navigation device paired. It improves especially when driving fast the communication within the vehicle clearly.

The level of vehicle interior noise rises very sharply at low frequencies, so that the language there is covered by the noise. To be possible through the DVE system to transmit little ambient noise, because this would make the interior noise level unnecessary increased, in a part of the methods described above, all frequencies each cut off according to speed below, for example, 200 to 500 Hz. The result is, that the basic speech frequency and the first multiples (harmonics) in transmitted signal are missing. The language thus sounds like a telephone, since it is typical Telephone network only allows sound transmission above 350 Hz.

In addition to using a hands-free phone, the procedure can also be used to Speech communication can be carried out inside the vehicle. However, it is Optimal sound quality is required to achieve acceptance among buyers.

In particular, in the methods that free speech from noise, e.g. B. spectral subtraction or coherence filtering, it happens that the variance of the Frequency components of noise in the order of the power of the Voice signal is coming. This means that effective noise suppression is no longer possible and the procedures used no longer apply.

The invention is therefore based on the technical problem arising from the prior art Technique known as well as the associated device for the reconstruction low-frequency speech components from medium-high frequency components to further develop and design that for a reproduction of the disturbed Speech signal a reproduction that is as close to nature as possible is made possible.

The technical problem outlined above is solved by a method with the features of claim 1 solved. First, above a cutoff frequency, at least two adjacent frequency components with increased amplitude in the speech signal certainly. After that, the fundamental frequency of the speech signal is called the frequency difference determined between the at least two adjacent frequency components. Finally becomes the low frequency with the help of the determined fundamental frequency and the speech signal Reconstructed frequency range below the cutoff frequency. The thus generated Synthetic speech signal can then be played back directly via a playback device be issued or saved for later transmission.

In other words, low-frequency signal components of the speech signal become synthetic generated, that is, reconstructed, and the remaining recorded speech signal added. The low-frequency speech components are reconstructed based on the unfiltered speech signals. This is exploited that the low-frequency speech parts accompanied by higher-frequency parts of the harmonics are so that the missing portions are estimated from the remaining signal to let.

In addition to the fundamental frequency, the frequencies of the harmonics of the fundamental frequency arranged below the limit frequency are determined and in addition to the fundamental frequency for a reconstruction of the low-frequency Frequency range used. Thus, the spectrally evaluated section of the speech signal the maximum information regarding the undisturbed speech signal exploited. The frequencies used for the reconstruction are marked with a respective spectral distribution and a predetermined amplitude to a synthetic Spectrum composed that the frequency range below the cutoff frequency in Voice signal corresponds. From this frequency segment and the speech signal above the reconstructed speech signal is then composed of the cutoff frequency. The low-frequency speech component therefore no longer has a noise signal since it is exclusively is composed of frequency components of the speech signal.

In a further embodiment of the invention, the low-frequency speech component can also can be determined directly from the speech signal. To do this, one of several band filters existing comb filter based on the fundamental frequency and the frequencies of the harmonics arranged below the cutoff frequency, the Frequency positions of the individual bandpass filters, the cut-off frequencies and the Correspond to harmonics. With the help of the comb filter, the speech signal is then in the Filtered range below the cutoff frequency, allowing the signal components to pass become part of the actual speech signal. In this way too Reconstruction of a largely undisturbed speech signal in the low-frequency range of the speech signal possible.

The decisive factor for the quality of the reconstruction of the low-frequency speech component is the Accuracy of the determined basic frequency of the speech signal. Since the Basic frequency continuously changed during speaking due to the sentence melody, a further improvement of the method is achieved in that at the beginning of a Speech contained speech section from the speech signal the fundamental frequency is determined and then this is adaptively tracked. Thus, in time The course of the speech signal determines the current fundamental frequency, so that the Reconstruction of the speech signal adapted as closely as possible to the course of the voice can be. An embodiment of such an adaptive tracking will continue explained in detail below.

In a further preferred manner, the amplitude of the at least one is below the Cutoff frequency generated frequency signal depending on the amplitudes of the frequency signals analyzed above the cutoff frequency are determined. In further Typical amplitude profiles of speech signals can preferably be used Find application not only in the frequency components, but also in the Amplitude distribution of the frequency components as precise an adaptation as possible to achieve natural speech signal.

It is further preferred that the cutoff frequency is dependent on the noise level, ie is determined in particular by the size of the interference signal. So it is low Interference signal levels, for example, only necessary to reduce the speech signal component below 200 Hz reconstruct, while it is necessary that the speech signal at high interference signal levels to be reconstructed in the frequency range below 500 Hz. When using the Procedure in a moving motor vehicle, the cutoff frequency can also be in Depending on the driving speed can be determined.

Another development is that the speech signal before a Conversion is subjected to interference signal exemption. The conventional methods known from the prior art are used, to pretreat the speech signal. The speech parts occur then more clearly in the spectrum and can be recognized more clearly and therefore more precisely and be reconstructed.

One application of the method described above is in one to play recorded voice signals while the motor vehicle is moving, thereby providing a reproduce as natural a language impression as possible.

Another application of the method according to the invention is to use a to reproduce a voice signal transmitted on a telephone connection. The basis lying problem is that the voice signals on telephone connections in the frequency range below 350 Hz contain no information. Therefore, for a lifelike reproduction of the speech signal from the low-frequency speech portion the frequency range above 350 Hz can be reconstructed. This can be done in carried out particularly advantageously by the method according to the invention become.

Fig. 1

eine spektrale Innengeräuschverteilung in einem fahrenden Kraftfahrzeug für unterschiedliche Fahrgeschwindigkeiten,

Fig. 2

ein Spektrogramm eines im tieffrequenten Bereich von einem Störsignal überlagerten Sprachsignals,

Fig. 3

ein Spektrogramm des in Fig. 2 dargestellten Sprachsignals ohne Störsignal,

Fig. 4

ein Spektrogramm des in Fig. 3 dargestellten Sprachsignals ohne Frenquenzanteile unterhalb der Grenzfrequenz von ca. 400 Hz,

Fig. 5

ein Spektrogramm der im Spektralbereich unterhalb der Grenzfrequenz von ca. 400 Hz rekonstruierten Sprachanteile,

Fig. 6

das vollständige rekonstruierte Sprachsignal entsprechend dem in Fig. 3 dargestellten Sprachsignal ohne Störsignalanteil,

Fig. 7

ein Blockschaltbild eines Ausführungsbeispiels einer erfindungsgemäßen Vorrichtung zur Rekonstruktion tieffrequenter Sprachanteile aus mittelhohen Frequenzanteilen,

Fig. 8

eine Einrichtung zur adaptiven Nachführung der Grundfrequenz und

Fig. 9

die spektrale Verteilung der Kennlinien der Bandfilter des Regelelementes zum Feststellen der frequenzabhängigen Leistungsverteilung im Mischspektrum in Bezug auf die feststehende Mischungsfrequenz von 2000 Hz.

According to a further teaching of the present invention, the technical problem described above is also solved by a device with the features of claim 12, while advantageous embodiments are specified in claims 13 to 16. The device and the method carried out with it are explained in more detail below on the basis of exemplary embodiments, reference being made to the attached drawing. Show in the drawing

Fig. 1

a spectral interior noise distribution in a moving motor vehicle for different driving speeds,

Fig. 2

a spectrogram of a speech signal superimposed in the low-frequency range by an interference signal,

Fig. 3

2 shows a spectrogram of the speech signal shown in FIG. 2 without an interference signal,

Fig. 4

3 shows a spectrogram of the speech signal shown in FIG. 3 without frequency components below the cut-off frequency of approximately 400 Hz,

Fig. 5

a spectrogram of the speech components reconstructed in the spectral range below the cut-off frequency of approx. 400 Hz,

Fig. 6

the complete reconstructed speech signal corresponding to the speech signal shown in FIG. 3 without any interference signal component,

Fig. 7

2 shows a block diagram of an exemplary embodiment of a device according to the invention for the reconstruction of low-frequency speech components from medium-high frequency components,

Fig. 8

a device for adaptive tracking of the fundamental frequency and

Fig. 9

the spectral distribution of the characteristics of the bandpass filter of the control element for determining the frequency-dependent power distribution in the mixed spectrum in relation to the fixed mixing frequency of 2000 Hz.

1 and 2, the starting point of the present invention is shown.

Fig. 1 shows a frequency-amplitude diagram of the interior noise level in one moving motor vehicle for different speeds between 60 km / h and 160 km / h. In this representation it is striking that especially at low frequencies below approx. 500 Hz the interior noise level compared to the others Frequencies of the interior noise signal increases sharply. However, since with normal Voice position the basic frequency and the first harmonics to the basic frequency in Frequency range below 1000 Hz and especially below 500 Hz is one Determination, i.e. a filtering out of the speech signal from the Interior noise signal considerably more difficult.

Fig. 2 shows a speech signal which has been superimposed by a background signal in a time-frequency representation as a spectrogram. This spectrogram is e.g. obtained by a Fourier transform (FFT) from a microphone signal. In Fig. 2 indicate different gray values of the individual segments of the spectrogram different intensities. On the one hand you can clearly see the increasing intensity (lighter gray values) in the range of low frequencies towards the value zero and on the other hand narrow-band frequency components that largely parallel to each other short periods of time pass. These latter narrow-band frequency components represent harmonics of the fundamental frequency of the corresponding speech signal, the - As described below - are evaluated according to the invention.

Fig. 3 shows a spectrogram of the speech signal shown in Fig. 2 without the Background noise, so that even the low-frequency speech components as narrow-band Frequency components can be seen in the spectrogram below 500 Hz. This Parts of the language have to be reconstructed.

Fig. 4 also shows the speech signal shown above, in which the speech components are cut off below a cut-off frequency of approx. 400 Hz. Such a signal corresponds approximately to the voice signal as it is transmitted on a telephone connection becomes.

5 shows an example of a reconstructed speech signal in the area below the Cutoff frequency of approximately 400 Hz and Fig. 6 shows the composite reconstructed Speech signal from the reconstructed speech portion shown in FIG. 5 and the in Fig. 4 frequency portion shown above the cutoff frequency of the original Spectrum. How the reconstructed language components are obtained is described below described in detail with reference to FIGS. 7 to 9.

7 shows in a block diagram a device for the reconstruction of low-frequency speech components from medium-high frequency components. The speech signal is _fed to a means 4 for determining frequency components ω _fa1 , ω _fa2 , ... of maxima in the speech signal above a predetermined cut-off frequency ω ₀ . For this purpose, the speech signal is first passed through a bandpass filter 6, so that only the frequency components between the cut-off frequency ω ₀ and a further frequency ω _{1 are} cut out and passed on for further processing. ω ₀ is for example in the range from 200 to 500 Hz, in particular 350 Hz, while the frequency ω _{1 is} in the range of 800 Hz, for example. The frequency section of the speech signal filtered out in this way is mixed in the mixing element 8, so that the sum and difference frequencies of the frequency components contained in the cut out section of the speech signal are formed. The difference frequencies are of interest, so that the signal emerging from the mixing element 8 is processed by means of a low-pass filter, so that only frequency components below an adjustable frequency ω _{2 are} let through. The smallest difference frequency can thus be determined, which corresponds to the distance between two spectral components arranged adjacent to one another in the speech signal. Since these are two harmonics of the fundamental frequency, the difference frequency represents the fundamental frequency ω _g . This fundamental frequency is then fed to means 12 for the reconstruction of the speech signal. The voice signal is fed via a further input of the means 12 via a delay stage 14 and a low-pass filter 16. Thus the means 12 have both the value of the fundamental frequency ωg and a predetermined frequency section of the speech signal available for a reconstruction of the signal containing the speech. The delay stage 14 is used to compensate for the time span Δt that is required for determining the fundamental frequency ω _g and the low-pass filter 16 is used to sensibly reduce the amount of data that is fed to the means 12 for the reconstruction of the speech signal.

The means 12 for the reconstruction of the speech signal below the cut-off frequency ω ₀ has two alternative methods in terms of circuitry.

As a first alternative, the fundamental frequency ω _{g is} used to generate a signal in the reconstructed speech signal that corresponds to the fundamental tone of the speech. In addition, the frequencies of the harmonics to the fundamental frequency ω _{g can be} determined by simply multiplying by the numbers N = 2, 3, 4, ..., so that in addition to the fundamental frequency ω _g for a reconstruction of the speech component below the limit frequency ω ₀ the frequencies ω _h1 , ω _h2 , ... of the first, second and further harmonics arranged below the limit frequency ω ₀ are used. The aim is to generate all the harmonics in the frequency section of the speech signal to be reconstructed, ie to simulate them. For a spectral distribution around each of these frequencies, an approximation is assumed to be a Gaussian distribution or another possible spectral distribution that can be defined via a half width and an amplitude. As a result, the spectral sections shown in FIG. 5 can be generated in the spectrogram, which are not or only partially recognizable in the noisy signal shown in FIG. 2.

As a further alternative for a reconstruction of the low-frequency speech component, there is the possibility that the means 12 have a comb filter which has a plurality of band filters, the spectral transmission functions of which are determined by the fundamental frequency ω _g and the frequencies ω _h1 , ω _h2 , ... . The spectral transmission function of each bandpass filter is also defined over a predetermined width, so that corresponding spectral sections are filtered out of the speech signal in the range of lower frequencies below the cutoff frequency ω ₀ . Since only the portions that contain the speech signal are filtered out of the spectrogram, the speech signal is reconstructed from the spectrogram. If noise suppression is additionally carried out, the background noise is also filtered out from the filtered signal components, so that an almost natural speech signal is generated.

As can also be seen in FIG. 7, the voice signal is delayed by a time difference .DELTA.t via a further delay stage 18 in order to allow adaptation to the time period necessary for the reconstruction of the low-frequency voice component. After passing through a high-pass filter 20 in which the speech signal is filtered out above the cut-off frequency ω ₀ , both this high-pass filtered signal and the reconstructed speech signal for frequencies ω <ω ₀ converge in the sum element 22, from which the reconstructed spectrogram shown in FIG. 6 is generated becomes. This spectrogram therefore consists on the one hand of the frequency component reconstructed below the cutoff frequency ω ₀ and of the original frequency spectrum above the cutoff frequency ω ₀ . After conversion into a loudspeaker signal, the spectrogram generated in this way leads to an almost natural-sounding speech reproduction.

As already explained above, the fundamental frequency ω _g generally does not remain constant in a speech signal due to the speech melody. It is therefore necessary to constantly redetermine the fundamental frequency ω _g . On the one hand, this can be done by continuously running through the previously described method, which was previously described with the aid of elements 4, 6, 8 and 10. On the other hand, however, more precise adaptive tracking of the fundamental frequency ω _g can be carried out. This is possible with a device which is shown in FIG. 8.

The fundamental frequency ω _{g, 0} initially determined at the beginning of a speech signal is multiplied to N times the value with the aid of a multiplication element 24. The (N-1) th harmonic to the fundamental frequency is thus calculated. The frequency of these harmonics is referred to below as the control harmonic and the associated frequency is denoted by ω _r .

The frequency ω _r is introduced into a control loop via a multi-port switch. In an initialization phase at the beginning of a word, the output of the multiplication element 24 is transferred from the multi-port switch 26 to the mixing element 28. After a short time - as described below - there is an estimated value ω _r , and the multi-port switch 26 is switched so that ω _r , is passed on to the mixing element 28.

The aim of the control loop is to determine the difference between the (N-1) th harmonic and a fixed frequency of, for example, ω _m = 2000 Hz. Ideally, ω _{r is} exactly the frequency of the (N-1) th harmonic. The mixing element 28 forms the difference between ω _r and ω _m . A sine wave generator generates a sinusoidal signal with the frequency _{that is} specified by its input signal ω _d . This is fed to a mixing element 32 which mixes the speech signal and this sinusoidal signal. After mixing has taken place, the mixed signal is output from the mixing element 32, which is fed to a control element 34 for determining the frequency-dependent power distribution in the mixed signal in relation to the fixed frequency ω _m .

Assuming that the frequency ω _{r of} the control harmonics supplied to the mixing element 28 exactly matches a harmonic in the current speech signal, the sum of the difference frequency ω _d , which was generated by the difference with the fixed mixing frequency ω _m and ω _r , and a frequency component of the speech signal that corresponds to the control harmonics, exactly the mixing frequency ω _m . This is reflected in a performance distribution (P distribution) in the performance spectrum. The power distribution will be maximum at the mixing frequency ω _m .

However, if the frequency ω _{r of} the control harmonic does not correspond to the current frequency of the corresponding harmonic in the speech signal, the power distribution will not reach its maximum at the frequency ω _m , but at a position shifted by a difference value Δω. A correction value for Δω can thus be determined, which is added to the current value of the frequency ω _{r of} the control harmonic. This results in the new value of the frequency ω _{r, new} , which is fed to the control loop again via the multiport switch 26. Subsequently, mixing takes place again in the mixing element 28 with a subsequent control sequence, as described above. If the basic frequency and thus also the frequency of the corresponding harmonics in the speech signal thus change in the course of the speech signal, this is compensated for by the control loop, so that a current value ω _r, which largely corresponds to the basic frequency ω _r , is continuously generated.

9 shows the characteristic curves of a plurality of band filters, which are provided for determining the power distribution in the control element 34. 9 results in a number of 7 band filters which are arranged around the fixed mixing frequency ω _m = 2000 Hz. If, for example, the maximum power falls within the pass band of the middle band filter, the correction value Δω = 0 is set. If, on the other hand, the maximum lies in one of the adjacent band filters, a corresponding correction value Δω ≠ 0 is generated in order to shift the maximum of the spectral power distribution into the pass band of the middle band filter with continued control.

The value ω _r is branched off from the control loop and output via a multiplication element 38, in which the current frequency ω _{r is} acted on by the factor 1 / N in order to generate the value of the fundamental frequency ω _{g, adapt} . The value of the fundamental frequency ω _{g is} thus continuously adaptively tracked, as a result of which the reconstruction of the low-frequency speech component from the medium-high frequency components is improved and brought closer to a natural speech signal.

Claims (16)

Process for the reconstruction of low-frequency speech components from medium-high frequency components, at which above a limit frequency (ω ₀ ) at least two adjacent frequency components (ω _fa1 , ω _fa2 , ...) are determined with increased amplitude in the speech signal and at which the fundamental frequency (ω _g ) of the speech signal is determined as the frequency difference between the at least two adjacent frequency components (ω _fa1 , ω _fa2 , ...) and in which the low-frequency range below the cut-off frequency (ω _g ) is reconstructed with the help of the determined fundamental frequency (ω _g ) and the speech signal. Method according to Claim 1, in which from the fundamental frequency (ω _g ) the frequencies (ω _h1 , ω _h2, ...) of the harmonics of the fundamental frequency (ω _g ) arranged below the limiting frequency (ω ₀ ) are determined and in addition to the fundamental frequency (ω _g ) be used for the reconstruction of the low-frequency range. Method according to Claim 1, in which, with the aid of a comb filter having a plurality of band filters, on the basis of the fundamental frequency (ω _g ) and the frequencies of the harmonics arranged below the limit frequency (ω ₀ ), the frequency positions of the band filters are set up, with the aid of which the speech signal in the range is filtered below the cutoff frequency (ω ₀ ). Method according to one of claims 1 to 3, wherein said contained at the beginning of a speech speech portion from the speech signal, the fundamental frequency (ω _g) is determined, and then the fundamental frequency (ω _g) is tracked adaptively. Method according to claim 4, in which calculated for adaptive readjustment of the fundamental frequency (ω _g) from the current value of the fundamental frequency (ω _g) the frequency (ω _r) of a control harmonic as the N-th harmonic, at which the difference between the frequency (ω _r ) of the control harmonic and a fixed mixing frequency (ω _m ) is formed, in which a sinusoidal signal (sin (ω _d )) _is generated with the difference or sum frequency (ω _d ) resulting from the difference formation, in which the sinusoidal signal (sin (ω _d )) is mixed with the speech signal and a mixed signal is generated, in which the frequency-dependent power distribution in relation to the fixed mixing frequency (ω _m ) is determined in the mixed signal, in which a correction value (Δω) for the frequency (ω _r ) of the control harmonic is calculated from the power distribution, in which the frequency (ω _r ) of the control harmonic is changed by the correction value (Δω) and fed to a renewed mixing with the fixed mixing frequency (ω _m ) and at which the fundamental frequency (ω _g ) is output, which corresponds to the corresponding fraction 1 / N of the frequency (ω _r ). Method according to Claim 5, in which, for a determination, the power distribution, the mixed signal is fed to a plurality of band filters (BF _n ) which cover adjacent frequency ranges centered around the fixed mixing frequency, Method according to one of claims 1 to 6, wherein the amplitude of the at least one frequency signal generated below the cut-off frequency in Dependence on the amplitudes of those analyzed above the cutoff frequency Frequency signals is determined. Method according to one of claims 1 to 7, wherein the cut-off frequency in Dependence on the noise level is determined. Method according to one of claims 1 to 8, wherein the speech signal before a Conversion to a spectrogram is subjected to an interference signal exemption. Application of a method according to one of claims 1 to 9 for the Playback of a recorded in a moving motor vehicle Speech signal. Application of a method according to one of claims 1 to 9 for the Playback of a voice signal transmitted by means of a telephone connection becomes. Device for the reconstruction of low-frequency speech components from medium-high frequency components, in particular for carrying out a method according to one of Claims 1 to 11, with means (4) for determining frequency components (ω _fa1 , ω _fa2 , ...) of maxima in the speech signal above a predetermined cutoff frequency (ω ₀ ), with means (8) for mixing the frequency components (ω _fa1 , ω _fa2 , ...) to determine the fundamental frequency (ω _g ) of the speech signal as the difference frequency between two adjacent frequency components (ω _fa1 , ω _fa2 , ...) and with means (12) for reconstructing the speech signal below the cut-off frequency (ω ₀ ) from the determined fundamental frequency (ω _g ) and the speech signal. Apparatus according to claim 12, characterized in that the means (12) for the reconstruction of the speech signal below the cut-off frequency (ω ₀ ) the spectrogram from the fundamental frequency (ω _g ) and the frequencies (ω _h1 , ω _h2 , ...) of the below the limit frequency (ω ₀ ) arranged harmonics of the fundamental frequency (ω _g ) with a predetermined spectral distribution and a predetermined amplitude distribution. Apparatus according to claim 12, characterized in that the means (12) comprise a comb filter with a plurality of band filters, the frequencies of the band filters being based on the fundamental frequency (ω _g ) and possibly one or more below the cut-off frequency (ω ₀ ) arranged harmonic of the fundamental frequency (ω _g ) are adjustable. Device according to one of claims 12 to 14, characterized in that there are provided for adaptive tracking of the fundamental frequency (ω _g ), a multiplier element (24) for generating the Nth harmonic of the fundamental frequency as the frequency (ω _r ) of a control harmonic, a mixing element (28) for mixing the frequency (ω _r ) of the control harmonics with a fixed mixing frequency (ω _m ), a sine generator (30) for mixing the difference or sum frequency (ω _d ) resulting from the mixture, a mixing element (32) for mixing the sinusoidal signal (sin (ω _d )) with the speech signal and for generating a mixed signal is generated, a control element (34) for determining the frequency-dependent power distribution in the mixed signal in relation to the fixed mixing frequency (ω _m ) and for calculating a correction value (Δω) for the frequency (ω _r ) of the control harmonics from the power distribution, a mixing element (36) for changing the frequency (ω _r ) of the control harmonics by the correction value (Δω) and with a multiplier element (38) for calculating the fraction 1 / N of the frequency (ω _r ) as the fundamental frequency (ω _g ). Apparatus according to claim 15, characterized in that the control element (34) has a plurality of band filters which cover frequency ranges lying next to one another centrally to the mixing frequency (ω _m ).

Images