RU2477533C2 - Method for multichannel adaptive suppression of acoustic noise and concentrated interference and apparatus for realising said method - Google Patents

Abstract

FIELD: radio engineering, communication.

SUBSTANCE: apparatus for suppressing acoustic noise and concentrated acoustic interference comprises: a unit for generating frame-segments, a unit for calculating spectral component levels, a unit for calculating the modulus of spectral components, a unit for calculating the control signal, a band-pass filter, a unit for generating a signal u_f(n), a unit for calculating dispersion, a unit for calculating threshold, a unit for multiplying spectral components with the control signal, an adder, a unit for calculating the inverse discrete Fourier transform and a low-pass filter.

EFFECT: high efficiency of transmitting information in telecommunication public address systems operating in conditions with acoustic noise and concentrated interference.

1 dwg

Description

The invention relates to devices for improving the efficiency of the operation of information transmission systems by acoustic speech signals operating under the influence of intense acoustic noise.

By auto No. 91490 [1] it is known a device for suppressing lumped interference, containing two parallel combs of bandpass filters, while the outputs of the first comb are connected to the inputs of the respective multipliers, the other inputs of which are connected to the outputs of the second filter comb through the threshold voltage generators, the outputs of the multipliers are connected to the inputs of the adder , from which the speech signal is fed to a band-pass filter from the output of which enters the communication channel. As a disadvantage of this device, it can be noted that the device does not suppress acoustic noise interference, which also occurs in telecommunication systems, speakerphone.

The aim of the invention is to minimize the negative impact of acoustic noise on the functioning of telecommunication systems for transmitting information by an acoustic speech signal, such systems as, for example, speakerphone systems (GHS).

It is known [2, 3] that GHS telecommunication systems operate, as a rule, under the influence of intense acoustic noise, which leads to loss of information. This problem is especially relevant when applying GHS systems to multifunctional sea vessels with a large number of subscriber stations that operate under conditions of intense exposure to acoustic noise interference and acoustic concentrated interference.

Acoustic noise interference is acoustic noise, the frequency spectrum of which, as a rule, is dispersed in the low-frequency part of the sound range. Acoustic concentrated noise is narrow-band tonal acoustic noise, the spectrum of which is concentrated in a narrow frequency band, and is almost harmonic oscillation. Such interference may occur at any frequency in the audio range.

Therefore, the aim of the invention is to maximize the attenuation of the influence of acoustic noise on the loss of useful information, that is, to obtain the maximum ratio P _c / P _w at the output of the GGS system operating under the influence of acoustic noise. Unlike the utility model described by Auth. No. 91490, using the device of this invention provides the suppression of both acoustic concentrated interference and acoustic noise interference.

Known acoustic noise suppression algorithms in acoustic signal information transmission systems have a structure based on linear filtering and include the nth number of high-order bandpass filters, which requires significant material resources and computational costs.

The essence of the invention lies in the fact that frames from N samples are formed from additive input signals (speech signals plus acoustic noise) by which the coefficients of the discrete Fourier transform (DFT) are calculated on a finite interval [4, 5], that is, the levels of components at frequencies are calculated l · f ₁ , where f ₁ is the frequency resolution interval when calculating the spectral function using the DFT; l is the number of the component.

The calculated levels of the l-th components are multiplied with the control signal from the output of the threshold device. The result of multiplication is fed to the multi-input adder.

The control signal of the l-th channel is formed according to the rule: if the module of the l-th spectral component does not exceed a certain threshold (the case of no interference), then the control signal is unity, the spectral component after multiplication is fed to the input of the adder; if the module of the lth spectral component exceeds the threshold k ₀ (the case of interference at a frequency l · f ₁ ), then the control signal will be zero. In this case, the control signals turn off S (l · f ₁ ), the module of which has a high level due to the presence at the frequency l · f _{1 of the} spectral component of acoustic noise. Thus, spectral components containing acoustic noise are excluded from the sum of the components of the total spectral function.

(9), сумматор (10), блок вычисления обратного ДПФ (11), фильтр нижних частот (12).Figure 1 shows a structural diagram of a device for suppressing acoustic noise. The device comprises a band-pass filter (1), a unit for generating frame segments (2), a unit for calculating the levels of spectral components (3), a unit for calculating a module for spectral components (4), a unit for computing a control signal (5), a unit for multiplying spectral components with a control signal (6), a signal conditioning unit u _f (n) (7), a dispersion calculation unit σ ² (8), a threshold calculation unit according to the algorithm

(9), adder (10), inverse DFT calculation unit (11), low-pass filter (12).

The proposed method is implemented as follows.

The input signal x (n), which is the additive sum of the speech signal and interference, is fed through the bandpass filter 300-3400 Hz (1) to the input of the frame-segmenting unit (2), while with the segment duration τ _s = 20 ms and the sampling frequency F _∂ = 10 kHz, the number N is 200 samples.

, ω₁=2πf₁, где f₁=F_∂/N - интервал дискретизации спектральной функции S(f) по частоте.In block (3), the levels of spectral components at frequencies l · f ₁ are calculated by the expression

, ω ₁ = 2πf ₁ , where f ₁ = F _∂ / N is the sampling interval of the spectral function S (f) in frequency.

The values of the numbers of the components l with f ₁ = 50 Hz for frequencies from 300 to 3400 Hz are in the range

where F _n - the lower frequency of the spectrum of the speech signal;

F _in - the upper frequency of the spectrum of the speech signal.

If a higher frequency resolution of the spectral function S (f) is needed, that is, it is necessary to reduce the sampling interval of the spectrum f ₁ for a fixed N, a new sequence L> N of the form

Thus, we obtain an addition of LN zero samples to a sequence of finite length from N samples, which allows us to achieve the necessary resolution f ₁ = F _∂ / L when calculating the spectral density using DFT on N samples of an analog signal [5].

Next, in block (4), the modules of the spectral components are calculated

,

,

Where

,

,

.

.

The value of the module Y ₁ is fed to the first input of block (5), in which it is used to calculate the control signal y _l (k ₀ ) according to the rule

To calculate the threshold level k _{0, the} signal from the output of the band-pass filter (1) is also fed to block (7), in which the signal u _f (n) is generated according to the rule

From the output of block (7), the signal u _f (n) is fed to the input of block (8), in which the dispersion value is calculated by the formula

,

,

для речевых сигналов [6].in this expression it is taken into account that

for speech signals [6].

From the output (8), the signal is fed to the input of the block (9), in which the threshold level value k ₀ = kσ is calculated. The threshold level value is supplied to the second input of the control signal calculation unit (5).

The signal from the output of the level calculation unit of the l-th spectral component (3) is fed to the input of the block (6), in which this spectral component S (l · f ₁ ) is multiplied by the control signal y _l (k ₀ ), taken from the output of the block (5). Thus, blocks (3), (4), (5) and (6) form the l-th channel. The output signals of the channels S (l · f ₁ ) · y _l (k ₀ ) go to the adder (10), the expression of the total signal at the output of the adder has the form

,

,

accordingly, in this sum there are no terms of l channels that are affected by acoustic noise.

The total signal S '(l · f _l ) is fed to the input of the inverse DFT block (11), from the output of the block (11) it is fed to the low-pass filter (12) with an upper cutoff frequency of 3400 Hz, from the output of the low-pass filter (12) receive speech filtered out from interference signal.

Thus, the described method for implementing the speech signal processing algorithm allows to increase the efficiency of information transfer in telecommunication systems of loud-speaking communication, operating under the influence of acoustic noise and concentrated noise.

Information sources

1. RF patent No. 91490 U1, IPC H04B 1/10. Publ. 02/10/2010. Bull. Number 32.

2. Handbook of technical acoustics. / Ed. M.Hekla and H.A. Muller. - L., Shipbuilding, 1980 .-- 440 p.

3. Babkin V.V. Noise reducing device for vocoder. // 9th Int. Conf. and the exhibition “Digital Signal Processing and its Application” (DSPA-2007). March 28-30, 2007, Moscow.

4. Rabiner LR, Gould B. Theory and application of digital signal processing. - M .: Mir, 1975 .-- 835 p., Ill.

5. Kropotov Yu.A. Research methods of spectral analysis of speech signals. / Yu.A. Kropotov., A.A. Bykov, A.Yu. Proskuryakov // Proceedings of the 18th International Crimean Conference "Microwave & Telecommunication Technology". Sevastopol, Ukraine. 2008. - V.1. - P.308-309. ISBN 978-966-335-167-4.

6. Kropotov Yu.A. Statistical parameters of signals in the design of operational-command telecommunication systems // In the world of scientific discoveries. - 2010 .-- N 6.1 (12). - S. 39-43.

Claims (1)

A device for suppressing acoustic noise and acoustic concentrated interference, comprising a band-pass filter, the output of which through a series-connected signal conditioning unit u _f (n), a dispersion forming unit and a threshold calculation unit is connected to the first input of the control signal calculation unit of each of l parallel channels, also an output the band-pass filter is connected to the block forming the frame segments, the output of which is parallel connected to the inputs l of the parallel channels, where l = 6 ... 68, the inputs of which are the inputs of the blocks calculating the levels of spectral components, the output of each of the l-th channel being connected to the input of the corresponding unit for calculating the module of spectral components, which is connected to the second input of the corresponding units for computing the control signal, the first input of which is connected to the output of the threshold level calculation units, while the outputs of the block calculating the levels of spectral components and the control signal computing unit of each l-th channel are connected to the corresponding inputs of the multiplication unit, the output of which is the output of the l-th channel, all l outputs of which are fed to the adder, from which the total spectral function is supplied to the inverse DFT calculation unit, from the output of the inverse DFT calculation unit, the restored signal is fed to the low-pass filter, from the output of the low-pass filter, the signal cleared from interference, enters the communication channel.