JPS59206900A - Noise remover - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a noise removal device used to remove white noise, etc. superimposed on a transmission signal in an audio signal transmission system such as a radio broadcasting system or a tape recorder. In particular, the present invention relates to a noise removal device that improves the S/N through mathematical calculation processing.

[Background technology and its problems]

In general, as a method for improving the S/N by removing noise components superimposed on a transmission signal through mathematical calculation processing, a method of correlating transmission signals has been known. In this correlation method, a fast Fourier transform (so-called F F
It is known that by performing analysis using the T) method, the relative strength of the frequency component of the transmission signal can be detected in the squared form, thereby increasing the S/H improvement effect.

Here, the time-series signal f (t) such as an audio signal is expressed as a Fourier series: jft)-i' A-(2)S(ω.t+ρn)
It can be expressed by the first equation, and its autocorrelation function ρ(τ) is as follows. The calculation aimed at the second equation can be obtained by analyzing f (X) as =3-F Ti'T.

By the way, since a sine wave function must be used for FFT calculation, a large-scale processing unit is generally required.
It takes an extremely long time to perform calculations on a so-called microcomputer with low calculation performance. Therefore, the above F
Conventionally, it has not been possible to perform S/N improvement processing using FT analysis in real time using a microcomputer. In addition, the second equation showing the autocorrelation function ρ(τ) above has a coefficient An2 and contains the same frequency component as the original signal f (t) shown in the first equation, but is replaced by the coefficient An. This requires arithmetic processing to perform the change, and this arithmetic processing also takes time. In addition, the characteristics of either the transmission signal component or the noise component are extracted based on the parameters obtained by autocorrelation, and
S/N by subtracting the noise component from the transmitted signal component
There are methods to improve this, but this method requires more complex signal processing and requires accurate extraction of features over time, resulting in a large-scale processing process.

[Purpose of the invention]

4- Therefore, the present invention provides a noise removal device with a novel configuration that can perform mathematical calculation processing for removing noise components superimposed on a transmission signal in real time using an arithmetic processing device comparable to a microcomputer. It provides:

[Summary of the invention]

A noise removal device according to the present invention is characterized by having a basic configuration as shown in FIG. 1 in order to achieve the above-mentioned object.

In FIG. 1, the Walsh transform processing unit performs arithmetic processing to indicate the energy distribution ζ on the frequency axis of an input signal given as a waveform signal on the time axis. Further, the subtraction processing section calculates the average energy level of the antiformunt band of the Walsh-transformed input signal, and performs arithmetic processing of subtracting the average energy level from all energy bands on the frequency axis. Then, the inverse Walsh transform processing section performs arithmetic processing to reproduce a waveform signal on the time axis corresponding to the energy distribution on the frequency axis that has been subjected to the -F notation subtraction process.

In the noise removal device having the basic configuration shown in FIG. 1, for example, the energy distribution on the frequency axis obtained by performing Walsh transform WFT on the input signal shown in the waveform diagram of FIG. 2A that does not include noise is Figure 2B
When the above input signal is superimposed with noise and the signal shown in the waveform diagram of FIG. 3A is subjected to the Walsh transform WFT, the energy distribution becomes as shown in FIG. 3B, and its Antiformant band (waveform nl
Calculate the average energy level k of l~In2),
The energy distribution obtained by cutting off this average energy level k from the total energy band is as shown in Figure 3C, and when the corresponding waveform signal on the time axis is obtained by inverse Walsh transform, the energy distribution is as shown in Figure 3. As shown in the figure, it is possible to reproduce a signal from which noise components have been removed.WFT processing uses square wave approximation, whereas F'FT processing uses sine wave approximation.
Calculations can be executed at the following calculation speeds.

〔Example〕

Hereinafter, specific embodiments of the present invention (a noise removal device according to the present invention) will be described in detail with reference to the drawings.

In the embodiment shown in the block diagram of FIG. 4, the input audio signal supplied as a waveform signal on the time axis through the signal input terminal 1 is converted into an analog-to-digital signal (A, /-D).
The digital data quantized in the unit 2 and representing the signal waveform is subjected to mathematical calculation processing in the calculation processing unit 3ζ to remove noise components, and then converted into digital analog (
D'/A) The signal is converted into an analog signal by the converter 4, reproduced into a waveform signal on the time axis, and outputted from the signal output terminal 5.

The mathematical calculation process for removing the noise component is carried out using the first microphone [cocomputer 11] for data input/output control and the second microphone [computer 12] that actually performs the mathematical calculation process, and the first microphone The processing is executed by the arithmetic processing unit 4, which is composed of a first buffer memory 14, a second buffer memory 15, and an arithmetic memory 16, into which digital data is written/read and transferred via a data selector 13 whose operation is controlled by the computer 11. Do it in time 7
- It seems like it's going to be done.

Here, the above A, /I) conversion unit 2 converts the signal waveform of the input audio signal supplied via the signal input terminal (by the way, the sampling frequency fs according to the sampling theorem)
Sampling is performed at , and quantization is performed for each sampled value, and as shown in FIG.

In addition, each memo 'J14, 15゜16 of the arithmetic processing section 3
each has a capacity to store at least one frame of digital data. The first and second buffer memories IJ 14 and 15 are connected to a common attenuator I/S by controlling the switching of the data selector 13 for each frame section by the first microcomputer 11. using the first microcomputer 1
From 11 onwards, access is alternately made for each frame section.

The first microcomputer 11 performs basic control operations as shown in the flowchart of FIG. It takes in the digital data of the input audio signal quantized by the converter 2 and supplies the digital data to the 2-bit/A converter 4, that is, controls the input/output of data. The data input/output control by this microcomputer 11 is shown in FIG. Data writing WJ and reading R are performed alternately. That is, the first microcomputer 11, as shown in flowchart 1 in FIG.
When one piece of digital data is taken in from the A/D converter 2 and written to the buffer memory, the data is read from the buffer memory, the address is advanced by one, and the end of one frame section is checked. While writing and reading the data, when one frame period is completed, the data selector 13 is switched to access the other buffer memory.

The second microcomputer 12 is designed to perform arithmetic processing as shown in the flowchart of FIG.

That is, the second microcomputer 12 is designed to perform arithmetic processing as shown in flowchart 1 in FIG.

That is, the second microcomputer 12 stores digital data for one frame stored in the first or second buffer memory 14 or 15 that is currently being accessed by the first microcomputer 11. The content of the two-note calculation memory 16, that is, the digital data g(nl) for one frame section of the input audio signal, is taken into the calculation memory 16 from the buffer memory that does not exist, and (n) -wal(m-n)
, , , , , , 3rd 2N,O Walsh transform WFT that calculates the wave number intensity G(m)
Process. Note that in the third equation, wal(rn”
n) indicates the Walsh function. Regarding the wavenumber intensity G- obtained by the above Walsh transform processing, the average energy level 1 (1<-(Σ0(m))/m2-m of the m section (rr++~m2) of the antiformant band ...
・・・・・・・Song/4th formula % formula %)) is calculated, and the wave number intensity G 111101'I]) is obtained by subtracting the -1- average energy level from the total energy band.
-〇(m)-1ri ・・・・・・・・・
. . . Performs signal processing PROCESS to obtain the fifth formula % formula %]). Then, the wave number intensity G (Ili obtained by this signal processing) is subjected to inverse Walsh transform IWFT processing to remove noise components, and digiticle data g(n) is calculated. Then, this digiticle data g(n) is Move to the buffer memory.
The processing time allocation state is shown in FIG. 5C.

11- The end timing of the arithmetic processing by the second microcomputer 12 and the switching control timing of the data selector 13 by the first microcomputer 11 match, and the second microcomputer 12
buffer memory 14, second buffer memory 15 and ζ
By alternately performing the above-mentioned calculation processing on the stored digitized data for one frame I/-, the mathematical calculation processing for noise removal can be performed in real time.

That is, each frame section a shown in FIG. 5A4.

The digital data of b, c, . , the second microcomputer 12 performs arithmetic processing on the data in section a, and the data subjected to the arithmetic processing is read out from the first buffer memory 14 during writing of the data in section C and is sent to the D/A. It is output via the converter 4. Similarly, data in section 12 is subjected to arithmetic processing during section C, and is read out from second buffer memory 15 during section d.

〔Effect of the invention〕

As is clear from the description of the embodiments described above, the noise removal device according to the present invention removes noise components through mathematical calculation processing using Walsh transform and inverse Walsh transform using square wave approximation. The required time can be shortened, and the first microcomputer for data input/output control and the second microcomputer for arithmetic processing can be reduced.
By alternately accessing the two buffer memories of the microcomputer, the above mathematical operations can be performed in real time using a microcomputer with low computing power, and the desired purpose can be fully achieved. I can do it.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the basic configuration of a noise removal device according to the present invention, and FIGS. 2 and 3 are waveform diagrams for explaining the operating principle of the present invention. FIG. 4 is a block diagram showing a specific embodiment of the present invention. FIG. 5 shows the operation of the above embodiment.
It is. FIG. 6 is a flowchart 1 showing the operation of the first microcomputer in the above embodiment. 7th
The figure is also a flowchart showing the operation of the second microcomputer. 1・・・・・・・・・・・・・・・・・・ Signal input terminal 2・・・・・・・・・・・・・・・ A/D converter 3...・・・・・・・・・・・・・・・ Arithmetic processing unit 4・・・・・・・・・・・・・・・・・・ D/A
Conversion section 5...... Signal output terminals 11, 12... Microcomputer 13......・Data selector 14.15... Buffer memory 16...
・・・・・・・・・・・・ Arithmetic memory 15- 825- +←Set-up field--1 +-1! -1-1...,...21, 噌---1 Figure 6 1N opening ff59-206900 (7) Figure 7 procedural amendment (voluntary) June 9, 1980 Commissioner of the Japan Patent Office Waka Kazuo Sugi 2, Relationship with the invention title case Patent applicant address 6-7-35, Kitashina-yo, Tokyo Parts-Yo-ku Name (2)
18) Sony Corporation (Name) Representative: Norio Ohga 4, Agent: 105 Address: 11th Mori Building, 11th floor, 6-4 Toranomon 2-chome, Minato-ku, Tokyo (50B) 8266 (main) 6, "Detailed Description of the Invention" column (7-1) of the specification to be amended Description N 1-ρ (τl = -, j , E, , Ancos (
ω. τ)” is corrected as follows. (7-2) The statement "calculation time" on page 5, line 5 of the specification is corrected to "calculation time." (7-3) Description on page 6, line 14 of the specification [Antipolmant zone-1 is corrected to be "antipolmant zone." (7-4) The description "Signal input terminal" on the second line of page 9 of the specification is corrected to "Signal input terminal 1." (7-5) Description 1 from page 11, line 4 to line 6 of page 11 of the specification, that is,... ” to be deleted. (7-6) The statement "to shorten time" on page 14, line 8 of the specification is corrected to "to shorten time." ” ni

Claims (2)

[Claims] (1) Arithmetic processing means that performs Walsh transform processing to represent an input signal given as a waveform signal on the time axis as an energy distribution on the frequency axis, and an average energy level in the antiformant band of the input signal subjected to the Walsh transform processing. an arithmetic processing means for calculating and subtracting the average energy level from the total energy band on the frequency axis, and reproducing a waveform signal on the time axis corresponding to the energy distribution on the frequency axis subjected to the subtraction process. A noise removal device comprising arithmetic processing means for performing inverse Walsh transform processing. (2) An analog-digital converter that quantizes an input signal given as a waveform signal on the time axis to form digital data corresponding to the waveform of the input signal, and a mathematical an arithmetic processing unit that performs digital arithmetic processing; and a digital-to-analog conversion unit that converts digital data subjected to arithmetic processing in the arithmetic processing unit into an analog signal; A first buffer memory, a second buffer memory, and an arithmetic memory each having a capacity to store digital data corresponding to a waveform of an input signal to be quantized in a 171/-m section; It consists of a first microcomputer for output control and a second microcomputer for mathematical operation processing, and the first and second microcomputers control the first buffer memory and the second microcomputer. The digital data from the analog-to-digital converter is accessed alternately for each frame section and sent to the calculation memory for each frame section via the first buffer memory or the second buffer memory. The second microcomputer performs a Walsh transformation process on the digital data transferred to and stored in the arithmetic memory, in which the input signal given as a waveform signal on the time axis is expressed as an energy distribution on the frequency axis; Calculate the average energy level of the anti-Holman I band of the input signal that has been subjected to this Walsh transform process, and subtract the above energy level from the total energy band on the frequency axis indicated by -, and then calculate the above energy level on the frequency axis that has undergone the above subtraction process. Inverse Walsh transform processing is performed to reproduce the digital data of the waveform signal on the time axis corresponding to the above energy distribution, and the digital data obtained by performing mathematical calculation processing in the second microcomputer is converted into the digital data obtained by performing mathematical calculation processing on the first microcomputer. A noise removal device characterized in that the microcomputer supplies the noise to the digital-to-analog converter via the first buffer memory or the second buffer memory.