FI118022B - Signal processor for noise suppression system - has canceller for subtracting protected noise from frequency analyser signal - Google Patents

Abstract

The processor includes frequency analyser for inputting a signal containing a noise to perform frequency analysis, a signal detector and noise predictor for inputting the frequency-analysed signal to predict the noise of the signal portion on the basis of a past noise information. A canceller is provided for subtracting the predicted noise from the frequency-analysed signal, and signal compositor for performing signal composition on the noise-cancelled signal. The frequency-analysed output of the frequency analyser and a cancel coefficient signal are provided to cancel the noise component considering the cancel coefficient from the frequency-analysed output. The signal compositor composes the cancelled-output of the canceller.

Description

118022 I

•. ·· '··! '? Cr

The signal processing unit

The present invention relates to a signal processing device for effectively removing interference from a signal containing interference, such as a signal interfering with the interference.

Fig. 1 is a diagram showing a sketch of a prior art interference suppression system (JP patent application 63-500 543).

In Figure 1, the noise-added audio signal at the input terminal is divided by the channel divider 19 into a plurality of selected channels. Thereafter, the gain of these individual preprocessed speech channels 10 is controlled by the channel gain converter 21 to be transformed into a signal to be described later, so that the gain on the channels with low voice / noise ratio is reduced. The individual channels containing the preprocessed audio are then recombined in the channel combiner 26 to produce an interfered audio signal available at the output.

The individual channels containing the preprocessed audio are also fed to the channel energy estimator 20 which generates the energy envelope values for each channel. The post-processed sound is fed to the channel energy estimator 22.

The post-processed estimated channel energy is used by the background interference estimator 23 to determine the 20 to noise / noise ratio.

The channel SNR estimator 24 compares the estimator 23 background interference estimates with the estimator: * · estimator 20 with the channel energy estimate to generate an SNR estimate. The SNR estimate * * * is used to select a specific gain value from a channel gain table * ·, ··, containing experimentally predetermined gain values. The channel gain, · | ί, 25 bias controller 25 generates the individual channel gain * · values of the converted signal, under the control of the SNR estimate.

*** * However, such a previously known interference suppression system adjusts the gain to attenuate the interference without eliminating the interference, so that there is not enough v: interference free signal. For example, in W.M. Kush-

• M

A: 30 nen et al: '' The effects of subtracive-type speech enhancement / noise reduction.:. algorithms for parameter Estimation for improved recognition and coding in high noise, ICASSP89, Vol. 1, May 23, 1989, Glasgow, disclose commonly available interference suppression techniques.

M / It is an object of the present invention, as defined in the claim, to reliably provide a signal processing device for generating an interference free signal 2 · ν 1 1 8022: by solving a previously known interference suppression system problem.

Fig. 1 is a block diagram illustrating a prior art interference suppression system; Fig. 2 is a block diagram showing a signal processing device;

Fig. 3 is a schematic diagram of the spectrum and the capstroke in said device;

Fig. 4 is a diagram illustrating an interference prediction in said device; Figure 5 is a diagram illustrating a method of elimination over time based on said device;

Fig. 6 is a diagram illustrating a method of elimination over time based on said device;

Fig. 7 is a block diagram showing a second signal processing device; Fig. 8 is a block diagram showing a second signal processing device; .

Figure 9 is a diagram illustrating the elimination of the device of Figure 8; factor;

Fig. 10 is a block diagram showing a second signal processing device; Fig. 11 is a block diagram showing a second signal processing device; and; . Fig. 12 is a block diagram illustrating an implementation of a signal processing apparatus according to the present invention.

• · ·, ·· *. Fig. 2 is a block diagram showing a signal processing device.

• »25 Microphone 1 is usually exposed to noise such as motor sound • ♦ · *;]. * Ni, in addition to sound S. Thus, microphone 1 produces an audio signal mixed with * · * * interference (S + N).

The A / D converter 2 converts the audio signal into a digital signal with interference as an analog signal.

·· v ί 30 FFT Device (Fast Fourier Transform) 3 Examples of Frequency Analysis [·. the audio device performs a fast Fourier conversion to the interfered audio signal, which is then converted to a digital signal.

* · T The signal detecting device 45 detects a signal portion of a signal mixed with a Fourier transform. The device 45 includes, for example, *: ··: 35 with a cepstral analyzer 4 for cepstrian analysis of a Fourier transformed signal, and a signal detector 5 for detecting the signal portion using the cepstral thus analyzed. The term "kepster", which is derived from the term "spectrum," in this application is symbolized by ο (τ) and is obtained by Fourier inverse logarithm of a short-term spectrum S (co).

118022 3

5M

c (t) = Σ log | S ((om) | 2cos (tro m) m = 0 τ: η is a measure of time and x (time) is called "kefrence" obtained from the 10 words "frequency" (frequency) ).

The waveforms of Figure 3 are provided as an example. Figure 3 (a) is a transient spectrum and Figure 3 (b) is its wavelet. The signal detecting device 5 detects a signal portion of the interference portion using a beaker. For example, a method for detecting the peak value of 15 cepstres is known as a method of separating a signal portion using a cepster. This means that the method employs a peak detector 51 to detect the peak of the analyzed cepstrin, and a signal / noise detector 52 to extract the signal based on the peak value information thus detected. The letter P in Fig. 3 (b) indicates the peak value, and the portion at which the peak occurs is determined to be part of the audio signal 20. The peak is expressed, for example, by a predetermined set: the threshold and the peak are compared to the threshold. t •: *:, ·. · * The interference prediction device 6 supplies a Fourier transformed signal to it: V: with its interfering disturbances and predicts the trace of interference in the signal portion.

: ***: Based on disturbance information. For example, as shown in Figure 4, the axis X

• · · 25 represents the frequency, axis Y the sound level and axis Z the time. The data of p1 and p2 through pi * at frequency f1 is taken to predict the preceding pi. For example, the mean of the misalignments through the pi pi is predicted to be pi. Alternatively, as the portion of the audio signal continues, the pie is multiplied further by the damping factor. In the present embodiment, the interference predictor 6 predicts the interference in the signal portion by using the signal portion information signaled by the signal detector 45. For example, when a signal portion is detected, the device 45 predicts an interference signal 4 · “*: based on data from a recent interference portion ··· *. called at the point that starts at the signal part. It is also preferred that the interference suppression device 6 uses the signal detection device 45 to detect the signal portion * * 35 (interference portion) to accumulate past interference information. λ 118022 4

The eliminator 7 subtracts the interference predicted by the interference predictor 6 from the Fourier transformed signal to which the interference is mixed. For example, the time-domain elimination is performed in such a way that the predicted interference waveform (b) is subtracted from the interfering audio signal (a), as shown in FIG. 5, thus allowing only signal output (c). Also, as shown in Fig. 6, frequency elimination is performed by converting (a) the audio signal containing the interference i into a Fourier transform, then subtracting from the signal so formed the predicted interference spectrum (c), and inversely Fourier. is converted to obtain an interference free (e) audio signal. Of course, the non-signal portion 10 may be determined to be pure interference by generating a signal obtained by inverting the output signal of the FFT device 3, and in the non-signal portion invert directly summing to the output signal of the FFT device 3 to completely eliminate the interference.

The IFFT device 8, as an example of a signal combining device, inversely Fourier converts the interference free signal obtained from the eliminator 7.

The D / A converter 9 converts the interference free signal, which is a digital signal received by the IFFT device 8, into an analog signal. In Figure 2, f indicates that the interference free signal is an analog signal.

The speech detector 10 indicates which word is the interference free speech signal thus obtained.

. The operation of the device described above will be described below.

**! · [Microphone 1 delivers sound that is mixed with noise and • f produces an audio signal with mixed noise (S + N) (see Figure 2, a).

The A / D converter 2 converts the audio signal with its interferences, which is * * 25 analog signal, into a digital signal. The FFT device 3 performs a fast Fou- * · * * rier conversion to the audio signal mixed with the interference and then; * ·· v: Converted to a digital signal (see Figure 2, b).

The signal detecting device 45 detects a signal portion of a signal which has been mixed: and thus has been Fourier transformed. For example, the analyzer 4 of the cryptography: 30 performs a cryptographic analysis of the Fourier transformed signal. The signal detector 5 further detects the signal portion using the * ::: beaker thus analyzed (see Fig. 2, c). For example, device 5 detects the peak of the cepstral to detect a signal • · ** ·· *. i: The interference predictor 6 supplies a Fourier transformed signal to it; ·: ··· 35 to which the disturbance is mixed, takes data from pT and p2 through silicon at frequency f 1, and computes the midbrain from the disturbing portions via p1 pi, which are transformed into pi. Also i i.

5; 1,180,222; In the present embodiment, the interference predictor 6 predicts the interference in the signal portion (see FIG. 2, d) based on data from the immediate past interference portion when viewed from the point beginning at the signal portion when the signal is detected using signal portion information detected by the signal detector.

The eliminator 7 subtracts the interference predicted by the interference predictor 6 from the Fourier transformed signal to which the interference is mixed (see Fig. 2, e).

The IFFT device 8 Fourier inversely converts the interference free signal 10 obtained by the eliminator 7.

The D / A converter 9 converts the interference free audio signal, which is a digital signal received by the IFFT device 8, into an analog signal (see Fig. 2, f).

The sound detector 10 indicates which word is the interference eliminated audio signal thus obtained. Since the signal is free of interference, its detection rate becomes high.

The device interference detector 6 may be a device for predicting the interference component of the signal section based on simply past interference information without using the signal detected by the signal detecting device 45. For example, the device 6 20 simply predicts that the past interference will continue to signal. . ·. moiety.

• · ·

The apparatus may also be applicable to the processing of other signals which are affected by the interference • i · VI ', and need not be limited to the processing of the audio signal.

Further, the device, though implemented here programmatically using a computer · · · · · * 25 network, may also be implemented using a specific hardware-based circuit.

··· v * As described above, the signal detector detects a signal portion of a frequency-analyzed signal mixed with the interference, predicts the signal portion of the interference with past interference information, and reduces the predicted interference 30 of the interfered signal, thus allowing complete interference - ·, bean formation.

• · · · ;;; When the signal detection device 45 «· '·· ** is used in the interference predictor 6 to detect a signal as a trigger to predict signal part interference, the interference ♦: *: can be more accurately predicted so that the signal from which the interference is more effectively eliminated. construct.

• · 111822; 6: 't • ·; .:

By looking at the drawings, another device will then be explained.

Fig. 7 is a block diagram of a signal processing device. In Figure 7, the number 71 denotes a band splitter for distributing an audio signal including an interference, 5 for each frequency band, as an example of a frequency analyzer for frequency analyzing the signal, number 72 indicating an interference as will be described later, and numeral 74 indicates a band combining device 10 for combining voice as an example of a signal generating device for generating a signal.

The device described above will be described in detail hereinafter. The band splitter 71 is supplied with an audio which includes an interference / interference input signal and performs a band division into m channel frequency bands and supplies 15 to the interference predictor 72 and the eliminator 73. The interference suppressor 72 predicts the interference component for each signal tone, interference prediction is performed, for example, as previously described and illustrated in FIG. 4. The m-channel signal 20 is supplied to the eliminator 73 from the bandpass device 71 and the interference prediction device 72 eliminates the interference in which. subtracting the interference from each channel under the control of the elimination multiplication signal, «· ·" V and supplying them to the band combiner 74. This means that the elimination is performed by multiplying the predicted interference component by the elimination coefficient. As described previously and as shown in Figure 5. Also, as * · V; previously described and shown in Figure 6, an elimination with respect to frequency v is performed .The band combiner 74 combines an m-channel signal supplied from the eliminator 73. , for outgoing sound.

The operation of the signal processing device thus constructed will be described below.

• The audio including the interference / interference input signal is band-transmitted into m- * ·· channel signals by the band-splitter device 71 and its interference component is predicted for each channel by the interference predictor 72. The signal is band-distributed to the m-channel interference device 71 »M ka 35 ka is supplied to each channel from the interference prediction device 72. The interference elimination ratio at that time is suitably set to improve awareness in each I. · '! 118022. * s \ s 7 channels using the elimination multiplier signal. For example, awareness is improved such that when an audio signal is present, the elimination coefficient is set to al-j even if the interference occurs, so that the interference is not eliminated much. After that ; the interference free m-channel signal obtained by the eliminator 73 is combined with the band combiner 73 to form an outgoing audio signal.

As described above, according to the present embodiment, the interference elimination ratio of the elimination device 73 can be suitably set for each band through the input of the elimination factor, and the elimination factor is accurately selected according to tone 10, thereby allowing the generation of interference-suppressed output sound.

Fig. 8 is a block diagram of another signal processing device. I

Figure 8 is reserved for the same numbering for the same devices as in the embodiment of Figure 7. This means that the number 71 indicates the bandwidth divider,! 15, 72 indicates interference predictor, 73 indicates elimination device, and 74 indicates band combining device. The tone frequency detector 87 detects the audio frequency of the input audio / interference input signal and supplies it to the channel coefficient setting device 88. The audio frequency is determined by several methods as shown in Table 1 and indicates the presence / absence of sound and the nature of the audio.

Instead of the tone frequency detecting device 87, another one may be used; t _; *. a device for detecting the sound part. The elimination factor setting device 88 is set. "·] To operate in such a way that it sets the number of elimination coefficients m based on the sound frequency provided by the detector 87, and supplies it, i.e., - **! * '25 to the subtractor 73.

The operation of the signal processing device thus constructed will be explained v * hereafter.

• · ·: The audio containing the interference / interference input signal is bandaged into m-channel signals by bandwidth divider 71, and its interference component predicted *: 30 for each channel by interference predictor 72. From a signal bandwidth to m-channel bandwidth divider 71 , * produced for each channel from the interference predictor 72.

* * * «« · * • • • • • • • • * * ······ 4 * 8 118022:> TABLE 1

Classification Frequency Separation-Properties Method (1) Parallel-Performs decision processing of frequency cycles based on the majority, separated by a 6-class simple waveform peak detector.

Waveform (2) Data conversion Discard non-frequency pulse processing candidates from waveform data of various logical operations.

(3) Zero Offsets Aim the repeated waveform calculation for multiple zero offsets.

(4) Self-correlation The self-correlation function of the waveform of the sound and center cut allows the spectrum to be smoothed and peak cutting allows simplification of the operation.

(5, a) Modified self-correlation function for LPC analysis of ice; correlation between nodes and polarization of LPF and residual signal allow simplification of operation. '

B

Correlation (5, b) SIFT Algorithm After processing the waveform down sampling, performs LPC analysis and smoothes the spectrum with an inverse filter. Returns; by interpolation of the correlation function for accuracy over time.

• * · * * · • * - ·· * -— ““ ”“ * - ~ * · · (6) AMDF Indicates the average of the amplitude of the periodicity by the difference function (AMDF). Also allows separation of residual signal by AMDF.

·· * '——. - - - - - - - - - - ™ - * t * (7) Kepstri Separates the spectral envelope and the fine field by Fourier transform of the power spectral · .. logarithm.

V: III

• *: \ Spectrum- (8) Periodic Determines the processing of the fundamental frequency, ·, * 'of the spectrum by the histogram of the harmonic hemisphere and determines the frequency by means of a common divisor of the higher harmonics ·· ".

• · * • * • * • · ·

V

11 11 ** hetkellä aset "" ella "syö" "11 11 11 11 11 11 11 11 11 11 11 11 9 9: 80 9 9 9 11 11 11 11 11 11 11 11 9 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11:

That is, when the predicted interference component is represented by ai, the interfering signal bi and the elimination factor alpha. the output signal of the eliminator 73 becomes ci (bi - alpha x ai). And its elimination coefficient is determined by the information being based on the tone frequency detection device 87. That is, the tone frequency detection device 87 supplies an audio / interference input signal to it and detects the frequency of the tone. The elimination factor setting device 88 sets the elimination factors as shown in Figure 9. This means that in Fig. 9, the min-10 coefficients for each band are shown, where f0 - f3 indicates the entire band of the audio / interference input signal. f0 - f3 is divided into an m channel for setting the elimination factor. f1 to f2 indicate in particular the band containing audio obtained using the audio frequency. Thus, in the audio band, the elimination coefficient is set low (near zero) so as to eliminate the interference as little as possible, which causes an improvement in consciousness. This is because the human ear can hear the sound even though the sound is slightly disturbed. For non-sound bands f0 to f1 and f2 to f3, the elimination coefficient is set to 1 to eliminate the interference sufficiently. The elimination coefficient in Fig. 9 is used when it is certain that no sound is present and only the interference is considered to occur, and it is given a value of 1 to sufficiently eliminate the interference. For example, if no vowel is present at the peak frequency, the signal cannot be considered to be. . ·. audio signal, so the signal is considered to be interfering. It is preferable that the elution factor of · · · 2 V in Figure 9 (a) and (b) can be suitably changed.

• «·

The apparatus described is applicable not only to the audio signal but also to other signal processing.

· * ··· * The device described can also be implemented, although it is implemented here programmatically using a computer, using a special hardware-based circuit. |

As is evident from the description above, the signal processing device si-: 30 includes an interference predictor for predicting the interference component, eliminating: * ;;; a signal is input, and which eliminates the interference component by taking into account the frequency elimination ratio of the frequency * ··· * word parsing, and the signal combining device combining the output signal eliminated by the eliminator so that when the interference component eliminates noise, the interference component eliminates control · 11 8022: | 10, which at the same time allows for the elimination of the disorder and for the improvement of consciousness.

Referring to the drawings, another device will be explained.

Fig. 10 is a block diagram of a signal processing device. The device is constructed as shown in Figure 10. That is, the interference prediction section 101 predicts the interference by means of an audio / interference input signal and a control signal supplied by the sound detecting section 103, and supplies the predicted interference suppression section 102. The elimination section 102 eliminates the interference audio / interference input signal 10 and outputting the outgoing audio to the audio detecting portion 103. The voice detecting portion 103 detects the presence / absence of the actual audio at the audio output to obtain an audio-detected output signal, and supplies the audio-detected output signal as a control signal to the interference prediction section 101.

The operation of the above structure will be explained. The sound which overlaps the interference / interference input signal 15 is supplied to an elimination section 102, wherein the interference is eliminated by a predicted interference controlled by the interference prediction section 101 to produce an audio output. The audio / interference input signal from which the interference is eliminated by the elimination section 102 is supplied to the audio detection section 103, wherein the presence / absence of audio is detected to provide an audio-detected output-20 signal. Thus, the interference detection portion 101 operates such that the portion uses as a control signal an audio-detected output signal indicating the presence / absence of sound and input from the audio detection portion 103 to predict the audio / interference signal, and to input the signal. for the audio-signal elimination section '•: · ** 102.

Therefore, according to the signal processing device, the sound detection is performed by the signal previously eliminated, which is the sound / interference • · *: result, while allowing the exact presence / absence of sound to be detected: ·: Irrespective of disturbance. With such detection, interference prediction can also be accurately performed and the interference effectively eliminated from the audio / interference input signal, resulting in pure outgoing audio.

•. 'v

Referring now to Figure 11, another device will be described.

Figure 11 is a block diagram of a signal processing device. The device is constructed ·· *: * as shown in Figure 11. This means that the first elimation section 105 eliminates the prediction of the first interference prediction section 104; 35 interference sound / interference input signal, and supplying the interference free signal to the sound interference section 106, the second interference prediction section 107 and the second elimination section • · • f '118022; 108. The voice detecting portion 106 detects the presence / absence of a signal input from the first elimination section 105 to obtain an audio detected output signal, and outputs the audio detected output signal as a control signal to the first interference prediction section 104 and the second interference prediction section 107. an elimination section 105 for obtaining a sound output from the signal supplied by the elimination section 105. The first interference prediction section 104 and the second interference prediction section 107 both use a control signal from the sound detecting section 106 to predict interference in the audio / interference input signal and to predict interference in the signal supplied from the first elimination section 105, respectively. The second interference prediction section 107 then supplies the predicted result to the second elimination section 108, which in turn makes a sound output from the eliminated result.

The operation of the above structure will be explained. The sound overlapping the interference / interference input signal is supplied to the first elimination section 105 15, wherein the interference is eliminated by a predicted interference controlled by the first interference prediction section 104. The first outgoing tone previously interrupted by the first elimination section 105 is fed to further, being accurately eliminated by the second predicted interference, which is supplied from the second prediction section 107 to obtain an outgoing sound. Also, the first output sound from which the interference has previously been eliminated by the first elimination section 105 is supplied to the sound detection section 106, where presence / absence is detected, to provide an audio output: 1V output signal (control signal). The first interference prediction section 104 now uses a * 1 control signal indicative of the presence / absence of sound from the audio detection section 106 to predict interference of the audio / interference input signal, and supplies * a first interference prediction signal to the first elimination circuit 105.

♦ »· A: The second interference prediction section 107 further operates such that the section 107 employs a control signal indicating the presence / absence of sound input from the audio detector section 106 in a s-: T: way to further accurately predict the en-30 first outgoing audio signal the interference has previously been eliminated, the originated interference signal at the first elimination section 105, and supplies a second * · · predicted interference to the second elimination section 108. Thus, in this signal processing device, the presence / absence of sound can be accurately detected regardless of the interference. continue to accurately predict and eliminate first. 35 outgoing noise from which the interference has previously been eliminated, which allows for the elimination of unstable interference with a much lower level of • 1 · 8022 12 1 club.

Referring now to Figure 12, another embodiment of the present invention will be described.

Fig. 12 is a block diagram of a signal processing device of another embodiment of the present invention.

The device is constructed as shown in Fig. 12. That is, the FFT processing section 121 converts the incoming signal into a frequency domain signal and feeds the converted signal to the cepster peak detector portion 122, the interference-10 preamble portion 125, and the elimination section 126. The kepster peak detector portion 122 detects the detected cepstrin peak to the audio frequency estimation section 123. The audio frequency estimation section 123 estimates the audio frequency from the cepstrin peak value and supplies the audio frequency to the windowing section 124, which in turn 15 generates a window the predicted signal elimination section 126, which in turn processes the signal input from the FFT processing section 121 according to the predicted interference, and outputs the processed sig bell audio-20 frequency enhancement section 127. The audio-frequency enhancement section 127 performs audio-frequency enhancement processing on signals supplied from the window portion 124. and an elimination section 126, and outputs the processed result to the IFFT section 128, which, in turn, converts the signal into a time domain for output.

* · · * "· 'The operation of the above structure will be explained. The signal to the present device ** 25 is converted into a frequency domain signal by FFT processing * · · Part 121. The input signal converted to the frequency domain is indicated by its binary v *: for the peak value, the tachograph peak detecting portion 122, and further on: T: determining the tone frequency by the tone estimation portion 123. Then, based on the determined tone frequency, the window portion 124 forms a suitable window for • · 4 for the input signal converted to the frequency band, assigns the interference component to the frequency domain and supplies the interference component to the elimination section 126.

* · · Then, the elimination section 126 eliminates each frequency component. ! *. 35 accurately interferes in the frequency domain obtained from the interference prediction section 125, from a signal converted to a frequency domain signal that is * 118022 13 input from the FFT processing section 121, and outputs the interference eliminated signal to the audio frequency enhancement section 127. a frequency gain signal from the elimination section 126 guided by a window, to perform the voice enhancement obtained from the windowing section 5 124, performs the audio enhancement and supplies the audio highlighted signal to the IFFT processing section 128. Thereafter, the IFFT processing section 128 converts the time area for output signal.

Thus, in a signal processing device according to an embodiment of the present invention, the interference is eliminated from the signal overlapping the interference, and a tone frequency enhancement portion is used to emphasize the audio component, thereby permitting the acquisition of an audio signal of excellent cognition.

While it is clearly preferred that the window formed by the window portion 124 in the above embodiment represents a harmonic wave structure of the sound, the window 15 may be a comb filter and a low pass filter. It is also clear that the tone frequency enhancement section 127 can be implemented simply by a multiplication circuit.

As is evident from the above embodiments of Figure 12, a device according to the present invention which eliminates the error by converting a signal into a frequency domain, includes an audio frequency prediction device for voice frequency prediction, a window glazing device controlled by a voice frequency, an interference prediction device, to eliminate the control of the output signal of the interference predictor, and to increase the frequency of the eliminated output of the tone frequency enhancer, by using the window of the window device, thereby eliminating the interference from the interfering signal, further emphasizing the • · · audio component to provide a highly perceptible audio signal.

Those skilled in the art will further appreciate that the foregoing description is a preferred embodiment, and that many modifications and variations may be made to the invention as defined in the appended claims without departing from its scope.

♦ 1 1 ♦ ♦ • • 1 1 1 • 1 1 1 »» »» »» »» »» »

Claims (1)

1. Talsignalbehandlingsanordning, kännetecknat av att den inhäller: en FFT-behandlingsanordnings (121) signal, i flick signal stör-5 ningar och tai sammanblandats, för FFT-behandling; en anning (122) for detecting the top of the frame, med hjälp av flashing en of the analysis of the FFT-utsignal of the FFT-behandlingsanordning (121); en anordning (123) for estimating an image (120), for detecting an image (122) for detecting such an anode (122) for detecting it; i en fönstringsanordning (124), med hjälp av vilken en fönsterutgäng picture on a basis av en förutsagd talfrekvensutgäng i nämnda anordning (123) f estimating av talfrekvens; 15 en anordning (125) förutsägelse av stöming, hjälp av flashing en störning förutsägs i such FFT-behandlingsanordnings (121) FFT-utsignal; en elimineringsanordning (126), med hjälp av vilken den förutsagda störningsutgängen för anordningen (125) förutsägelse av störning eliminating such FFT-behandlingsanordnings FFT-utgäng; 20 en anordning (127) för betoning av talfrekvens, med hjälp av vilken. den den eliminineringsanordningen (126) eliminerade utsignalen concrete med hjälp av fönstringsanordningens (124) fönstringsutgäng; och * · * * y * en IFFT-behandlingsanordning (128), med hjälp av vilken den beto- * ·: ·; nade utsignalen som kommer frän nämnda anordning (127) för betoning av 1 25 talf frequency IFFT-behandlas. • · · • • • • * * • i · · * »· * · • *: **: · * I" ":! · * · • m * · 1 ··· j • ·! • * * · · ♦ # * · * · • · · * «• · * * *