CN105244040A - Audio signal consistency comparison method - Google Patents

Abstract

The present invention discloses an audio signal consistency comparison method which comprises the steps of obtaining sampling data, expanding a length, calculating a fast correlation similarity, calculating a similarity and an offset, judging that two paths of data are consistent if the similarity is larger than or equal to a threshold value, judging whether the sampling data is subjected to approximated envelope line processing or not if the similarity is smaller than the threshold value, judging that the two paths of data are inconsistent if the sampling data is subjected to approximated envelope line processing, and carrying out approximated envelope line processing on the sampling data if the sampling data is not subjected to approximated envelope line processing and then repeating the above steps. According to the scheme of the invention, the consistency of two paths of audio signals can be accurately judged, and the method is suitable for the monitoring of a broadcast signal by a radio station.

Description

The control methods of a kind of sound signal consistance

Technical field

The present invention relates to audio data analysis process field, especially relate to the control methods of a kind of sound signal consistance.

Background technology

In whole transmission in broadcasting station, monitoring link, be all carry out detecting for the physical index of audio frequency, such as: whether break, level is on the low side or higher, anti-phase etc.But whether correctly detection means is lacked for the content of audio frequency, whether the signal that such as master control finally exports is the signal that studio's sound console exports, whether air-launched signal is that the problems such as the signal that exports of master control final stage do not have suitable detection means at present, to only have etc. to there occurs accident and just can detect.

State Intellectual Property Office of the People's Republic of China discloses on Dec 01st, 2010 patent documentation that publication number is CN101902677A, and title is audio detection device and method, and this device comprises: MIC electric circuit inspection module and phonation circuit detection module; According to amplitude and frequency, MIC electric circuit inspection module, for obtaining amplitude and the frequency of the sine wave signal after MIC circuit sampling, determines that whether MIC circuit is qualified, and sine wave output signal; Phonation circuit detection module comprises: change-over panel and the processor with MIC circuit; Change-over panel, for the simulating signal that phonation circuit of sampling exports, and is converted to digital signal by simulating signal, is sent to processor; Processor, for carrying out analyzing and processing to the amplitude of digital signal and frequency, judges that whether phonation circuit is qualified.This scheme cannot carry out conforming judgement to sound signal.

Summary of the invention

The present invention mainly solves the technical matters can not carrying out accurately judgement to sound signal consistance existing for prior art, a kind of characteristic signal analysis can passed through audio frequency is provided, detect the sound signal consistance control methods whether two paths of signals content is consistent, be convenient to broadcasting station Timeliness coverage and broadcast fault.

The present invention is directed to that above-mentioned technical matters mainly solved by following technical proposals: the control methods of a kind of sound signal consistance, comprises the following steps:

S01, from first via sound signal, obtain N number of sampled data as first via sampled data, from the second tunnel sound signal, obtain N number of sampled data as the second tunnel sampled data;

S02, by mending 0 after first via sampled data, to extend to length be 2N, and by mending 0 after the second tunnel sampled data, to extend to length be 2N;

S03, the first via sampled data after expansion is carried out to Fast Fourier Transform (FFT) and then asked conjugation to obtain first via data, then carry out Fast Fourier Transform (FFT) to the second tunnel sampled data after expansion asks conjugation to obtain the second circuit-switched data, ask the cross correlation function of first via data and the second circuit-switched data, carrying out Fast Fourier Transform (FFT) to cross correlation function, obtain similarity data;

S04, choose the maximal value of absolute value in similarity data, then calculate Similarity value and maximal value side-play amount;

If S05 Similarity value is more than or equal to consistance threshold value, then think that first via sampled data is consistent with the second tunnel sampled data, contrast flow process terminates; If similarity is less than consistance threshold value, then enter step S06;

S06, judge that whether first via sampled data and the second tunnel sampled data are through approximate envelop process, if do not had, then carry out approximate envelop process to first via sampled data and the second tunnel sampled data, then repeat step S02 to S05; If through approximate envelop process, then judge first via sampled data and the second tunnel sampled data inconsistent.

Said process more adequately can judge the similarity of first via sampled signal and the second tunnel sampled signal.According to the side-play amount calculated to after sound signal skew, the consistance of sound signal more adequately can be calculated.

As preferably, Similarity value ratio computing formula is as follows:

ratio＝fSample _max*M/sum

In formula, fSample _maxfor the maximal value of absolute value in similarity data, M is the number of similarity data, sum be the absolute value of all similarity datas and.

As preferably, maximal value side-play amount offset determines according to following methods: fSample _maxfor the data of i-th in similarity data, if i is less than or equal to M/2, then offset=i; If i is greater than M/2, then offset=i-M/2; FSample _maxfor the maximal value of absolute value in similarity data, M is the number of similarity data.

As preferably, approximate envelop process is specially: first via sampled data is divided into the window that length is 2 milliseconds, gets maximal value and form first window maximum data * X1 from each window; Second tunnel sampled data is divided into the window that length is 2 milliseconds, from each window, gets maximal value form Second Window maximum data * Y1; Calculate the average meanX1 of the absolute value of first via sampled data; Calculate the average meanY1 of the absolute value of the second tunnel sampled data; Each data in * X1 are deducted meanX1, and the data obtained are as first via sampled data; Each data in * Y1 are deducted meanY1, and the data obtained are as the second tunnel sampled data.

For volume smaller and make an uproar in the end obvious audio content, the similarity-rough set calculated directly carrying out step S02-S05 is little, generally can be less than consistance threshold value, can eliminate small-signal interference, improve differentiation rate after now carrying out approximate envelop process.

As preferably, the span of described consistance threshold value is 20-30.

By consistance threshold value accurately can be oriented after a certain amount of sampling analysis.

As preferably, when continuous three times judge that first via sampled data and the inconsistent then identification first via sound signal of the second tunnel sampled data and the second tunnel sound signal are inconsistent.

When there is first via sound signal and the inconsistent situation of the second tunnel sound signal, can carry out early warning, prompting staff carries out revising or safeguarding.

This programme may be used for following occasion:

Radio station master control:

For judging the consistance of master control final stage output signal and sound console output signal;

Master control final stage signal and empty consistance of to collect mail number;

Transmitting station:

Empty to collect mail number and the consistance of optical transmitter and receiver output signal;

Monitoring center:

User judges that whether broadcasting signal is disturbed or invade;

The substantial effect that the present invention brings is, reliably can calculate similarity and the side-play amount of two-way audio signal, eliminates the impact of sound signal decay, has high accuracy of judgement degree.

Accompanying drawing explanation

Fig. 1 is a kind of process flow diagram of the present invention.

Embodiment

Below by embodiment, and by reference to the accompanying drawings, technical scheme of the present invention is described in further detail.

Embodiment: a kind of sound signal consistance control methods of the present embodiment, as shown in Figure 1, comprises the following steps:

Record the voice data of certain sampled point number, first extension length is twice, is 0 by assignment below.

floatfSample[N*2]

fSample[i]＝0N≤i≤N*2

Then sampled data is carried out FFT conversion, fft algorithm is described below:

Strange according to discrete fourier transform, even, empty, the characteristics such as reality, improvement is carried out to the algorithm of discrete Fourier transform (DFT) and obtains fast Fourier transformation algorithm FFT, if x (n) is the sequence of complex numbers of N item, converted by DFT, the calculating of arbitrary X (m) all needs N complex multiplication and N-1 complex addition, and a complex multiplication equals four real multiplications and twice real addition, one time complex addition equals twice real addition, even if a complex multiplication and a complex addition are defined as once " computing " (four real multiplications and four real additions), so obtain the X (m) of N item sequence of complex numbers, namely N point DFT conversion approximately just needs N^2 computing.In FFT, utilize periodicity and the symmetry of WN, a N item sequence (is established N=2k, k is positive integer), be divided into the subsequence of two N/2 items, each N/2 point DFT converts needs (N/2) 2 computings, then the DFT of two N/2 points conversion is combined into the DFT conversion of a N point with N computing.After such conversion, total operation times just becomes N+2* (N/2) ^2=N+ (N^2)/2, saves the operand of about 50%.

$X\left(k\right)=sumx\left(n\right)*\exp (-j*2*\mathrm{\π}(k-1)*\frac{n-1}{n})$

1≤K≤N

$x\left(n\right)=\left(\frac{1}{N}\right)sumx\left(k\right)*\exp (j*2*\mathrm{\π}*(k-1)*\frac{n-1}{n})$

1≤n≤N

By the two-way sampled data after FFT budget, ask conjugation, then carry out a FFT computing, process is called calculating fast correlation similarity.

Then the data of Fourier transform are got maximal value, calculate maximum average value by number of samples, and calculate maximal value side-play amount by the start bit of maximal value, so just can calculate the sampled point side-play amount of Similarity value and two-way audio, computing method are:

Sampled point number: N

Side-play amount: offset

Similarity: ratio

Data after Fourier transform: fSample, because sound signal is sinusoidal wave, have positive and negative values, so absolute value should be used to carry out size contrast when calculating.

Data volume summation: sum

ratio＜|fSample[i]|？ratio＝|fSample[i]|offset＝i

sum＝sum+|fSample[i]|

1≤i≤N

$offset>\frac{N}{2}?offset=offset-N$

$ratio=\frac{ratio*N}{sum}$

After similarity (ratio) value is greater than certain value range (this value is commonly defined as 20-30 after long-term test), just two paths of signals can be regarded as consistent, and then according to the side-play amount calculated (offset), carry out the skew of two paths of signals, migration algorithm is:

fSample[i]＝fSample[i+offset]

1≤i≤(N-offset)

Follow-uply calculating is carried out to shifted data more will press close to contrasting region by Similarity value, improve accuracy.

Smaller for volume, the obvious audio content and make an uproar in the end, the approximate value drawn in current algorithm is smaller, generally can be less than predetermined scope, this value fluctuation situation is obvious, and Similarity value is generally at about 10-15, during this situation, adopt the mode of getting the approximate envelop of voice band line to carry out secondary calculating, small-signal interference can be eliminated and improve differentiation rate.Get the maximal value of 2 milliseconds of windows at every turn, obtain * X1, * Y1, and the average of two paths of signals:

Setting sampling rate is 48000, and so the window sample number of 2 milliseconds is 96

$X\[j\]=\underset{i\≤k\≤i}{\max }fSample\[k\]$

1≤i≤N；i+＝96；j++

sum＝sum+|fSample[i]|

1≤i≤N

$meanX=\frac{sum}{N}$

Average in * X1, * Y1 signal is removed:

rX＝*X1-meanX1

rY＝*Y1-meanY1

Then to data rX, the rY of secondary calculating again before step, calculate fast correlation similarity, the data calculated carried out again to the calculating of similarity and side-play amount, obtain new comparing result, improve accuracy further.The frequency occurred in actual environment due to this class audio frequency is not high, and twice calculating can promote CPU burden, therefore secondary authentication policy when being lower as Similarity value exists.

The basis of fast correlation similarity adds determination strategy again, and continuous 3 decision threshold are less than setting value (being generally 20-30), and think that two paths of signals is inconsistent during side-play amount fluctuation range comparatively large (being greater than the sampling of 1s).

After fast correlation Similarity measures, similarity and the side-play amount of two-way audio can be calculated reliably, again by after a certain amount of sampling analysis, just can orient decision threshold accurately, just can effectively detect the consistance of two paths of signals according to this threshold values, early warning, and owing to having carried out twice FFT computing, get the calculating that similarity is the signal curve carried out, the decay of sound signal does not affect result of calculation, and accuracy of judgement degree can reach more than 99.99%.

Specific embodiment described herein is only to the explanation for example of the present invention's spirit.Those skilled in the art can make various amendment or supplement or adopt similar mode to substitute to described specific embodiment, but can't depart from spirit of the present invention or surmount the scope that appended claims defines.

Although more employ the term such as sampled data, Fast Fourier Transform (FFT) herein, do not get rid of the possibility using other term.These terms are used to be only used to describe and explain essence of the present invention more easily; The restriction that they are construed to any one additional is all contrary with spirit of the present invention.

Claims (6)

1. the control methods of sound signal consistance, is characterized in that, comprises the following steps:

S01, from first via sound signal, obtain N number of sampled data as first via sampled data, from the second tunnel sound signal, obtain N number of sampled data as the second tunnel sampled data;

S02, by mending 0 after first via sampled data, to extend to length be 2N, and by mending 0 after the second tunnel sampled data, to extend to length be 2N;

S03, the first via sampled data after expansion is carried out to Fast Fourier Transform (FFT) and then asked conjugation to obtain first via data, then carry out Fast Fourier Transform (FFT) to the second tunnel sampled data after expansion asks conjugation to obtain the second circuit-switched data, ask the cross correlation function of first via data and the second circuit-switched data, carrying out Fast Fourier Transform (FFT) to cross correlation function, obtain similarity data;

S04, choose the maximal value of absolute value in similarity data, then calculate Similarity value and maximal value side-play amount;

If S05 Similarity value is more than or equal to consistance threshold value, then think that first via sampled data is consistent with the second tunnel sampled data, contrast flow process terminates; If similarity is less than consistance threshold value, then enter step S06;

S06, judge that whether first via sampled data and the second tunnel sampled data are through approximate envelop process, if do not had, then carry out approximate envelop process to first via sampled data and the second tunnel sampled data, then repeat step S02 to S05; If through approximate envelop process, then judge first via sampled data and the second tunnel sampled data inconsistent.

2. a kind of sound signal consistance according to claim 1 control methods, is characterized in that, Similarity value ratio computing formula is as follows:

ratio＝fSample _max*M/sum

In formula, fSample _maxfor the maximal value of absolute value in similarity data, M is the number of similarity data, sum be the absolute value of all similarity datas and.

3. a kind of sound signal consistance according to claim 1 control methods, is characterized in that, maximal value side-play amount offset determines according to following methods: fSample _maxfor the data of i-th in similarity data, if i is less than or equal to M/2, then offset=i; If i is greater than M/2, then offset=i-M/2; FSample _maxfor the maximal value of absolute value in similarity data, M is the number of similarity data.

4. a kind of sound signal consistance control methods according to Claims 2 or 3, it is characterized in that, approximate envelop process is specially: first via sampled data is divided into the window that length is 2 milliseconds, gets maximal value and form first window maximum data * X1 from each window; Second tunnel sampled data is divided into the window that length is 2 milliseconds, from each window, gets maximal value form Second Window maximum data * Y1; Calculate the average meanX1 of the absolute value of first via sampled data; Calculate the average meanY1 of the absolute value of the second tunnel sampled data; Each data in * X1 are deducted meanX1, and the data obtained are as first via sampled data; Each data in * Y1 are deducted meanY1, and the data obtained are as the second tunnel sampled data.

5. a kind of sound signal consistance according to claim 1 control methods, is characterized in that, the span of described consistance threshold value is 20-30.

6. a kind of sound signal consistance according to claim 1 control methods, is characterized in that, when continuous three times judge that first via sampled data and the inconsistent then identification first via sound signal of the second tunnel sampled data and the second tunnel sound signal are inconsistent.