CN106409302A - Audio frequency watermark method and system based on embedding area selection - Google Patents

Abstract

The present invention provides an audio watermarking method and system based on embedding area selection. The embedding process includes reading audio files, first judging whether each frame signal can be used as an embedding area, and then selecting the embedding frequency segment of the audio watermark; Discrete Fourier transform, generate a binary pseudo-random spread spectrum sequence, embed the watermark, and transform it to the time domain; the detection process includes reading the audio file to be detected, judging whether each frame signal can be used as an embedded area, and calculating the detection range The starting point and the end point of the frequency domain, the discrete Fourier transform is performed to generate a binary pseudo-random spread spectrum sequence, the sufficient statistics of the detection are calculated, and the detected watermark bits are obtained. The present invention proposes to filter out the transient signal by the ratio of the maximum energy to the minimum energy in the frame to improve the accuracy of watermark detection, and to improve the robustness of the watermark by embedding the watermark in a frequency band where the human ear perceives significantly.

Description

Audio watermarking method and system based on embedding region selection

technical field

The invention relates to the technical field of digital audio watermarking, in particular to an audio watermarking method and system based on embedding region selection.

Background technique

Digital audio watermarking is a signal processing operation that adds certain digital information to audio signals to achieve the purposes of document authenticity identification, copyright protection, and information hiding. The selection technology of the audio watermark embedding area refers to selecting the appropriate audio area to embed the watermark before embedding the watermark into the audio signal. The traditional audio watermarking technology does not consider the characteristics of the audio signal, and embeds the watermark on the entire audio file, which will cause 1) after the watermark is embedded in the area with low audio signal amplitude, the amplitude exceeds the masking threshold to generate noise, which destroys perception Transparency; 2) For transient signals with severe changes in the audio signal, the variance of the audio signal in this area is very large, and the watermark bit error rate is very high when the watermark is detected after embedding the watermark; 3) Embedding the watermark in the frequency domain, If the watermark is embedded in an area that is not obvious to the human ear, part of the watermark will be lost after signal processing or audio lossy compression, resulting in a high bit error rate for watermark detection.

Contents of the invention

The purpose of the present invention is to provide audio watermark technology for selecting a region to embed, so that the watermark can be embedded in a suitable audio region, avoiding unnecessary noise and reducing the occurrence of bit errors.

In order to achieve the above object, the technical solution provided by the present invention provides an audio watermarking method based on embedding region selection, including embedding process and detection process,

The embedding process includes the following steps,

Step A1, read the audio file, obtain the signal x _n of the time-domain audio of the nth frame after the sampling rate fs1 and the frame division, and the frame length is N,

First, judge whether each frame signal x _n can be used as an embedding area,

Then, for each frame signal x _n that can be used as an embedding area, the selection of the embedding frequency segment of the audio watermark is carried out, and the embedding frequency segment of the audio watermark is selected, and the start frequency of embedding preset according to the frequency part sensitive to the human ear is set to be FWMIN, the end frequency is FWMAX, the start embedding point freqmin1 and the embedding end point freqmax1 of a frame are calculated as follows,

freqmin1=floor((FWMIN×2.0/fs1)×N)

freqmax1=floor((FWMAX×2.0/fs1)×N)

Among them, floor is a function of rounding down;

Step A2, performing discrete Fourier transform on each frame signal x _n capable of embedding a watermark to obtain a frequency domain signal X _n ;

Step A3, using the key key as a random number seed to generate a binary pseudo-random spread spectrum sequence u with a length of freqmax1-freqmin1+1;

Step A4, according to the spread spectrum sequence u, the frequency domain signal _Xn and the watermark bit b, the watermark is embedded to obtain the frequency domain signal after the watermark is embedded, and the calculation is as follows,

|X′ _n |=|X _n |+bαu

Among them, α is a constant, which controls the embedding strength of the watermark, |X _n | and |X′ _n | represent the frequency domain amplitude before embedding the watermark and the frequency domain amplitude after embedding the watermark, and then get the embedding watermark by Euler's formula The frequency domain signal of

Among them, ∠X _n represents the phase of the frequency domain signal, X′ _n represents the frequency domain signal after embedding the watermark, and e is the mathematical natural exponent;

Step A5, transforming the frequency-domain signal X' _n embedded with the watermark into the time domain to generate an audio file embedded with the watermark;

The detection process includes the following steps,

Step B1, read the audio file to be detected, and obtain the nth frame signal z _n and sampling rate fs2 after time domain framing,

First, judge whether each frame signal x _n can be used as an embedding area;

For each frame signal x _n that can be used as an embedding area, as the signal to be detected, calculate the starting point freqmin2 of the detection range and the end point freqmax2 of the frequency domain

freqmin2=floor((FWMIN×2.0/fs2)×N)

freqmax2=floor((FWMAX×2.0/fs2)×N)

Step B2, perform discrete Fourier transform to obtain the frequency domain signal Z _n of the signal to be detected, and record the corresponding frequency domain amplitude value as |Z _n |;

Step B3, using the key key as a random number seed to generate a binary pseudo-random spread spectrum sequence u with a length of freqmax2-freqmin2+1;

Step B4, according to the spread spectrum sequence u and the frequency domain amplitude value |Zn| of the signal to be detected, calculate the sufficient statistic _{r n for} detection as follows,

If the sufficient statistic r _n ≥ 0, then the detected watermark bit is b=1; otherwise, the detected watermark bit is b=0.

Moreover, in step A1 and step B1, it is judged whether each frame signal x _n can be used as an embedding area, and the implementation method is as follows,

1) Average energy of signal x _n If the size exceeds the preset corresponding threshold τ ₁ , it is a silent zone and watermarks are not allowed to be embedded;

2) If the signal x _n contains a transient signal, it is not allowed to embed the watermark.

Moreover, whether the signal x _n contains a transient signal is judged by the following method,

Assuming that a frame signal is decomposed into S blocks, the energy of the S blocks is calculated respectively, and the energy ratio rate of the block with the largest energy and the block with the smallest energy is compared with the preset corresponding threshold τ ₂ , if the rate is greater than τ ₂ , it is considered that the Frame signals contain transient signals.

The present invention also correspondingly provides an audio watermark system based on embedding region selection, including an audio watermark embedding subsystem and a watermark detection subsystem,

The audio watermark embedding subsystem includes the following modules,

Select an appropriate area embedding module to read the audio file, and obtain the signal x _n of the time-domain audio of the nth frame after the sampling rate fs1 and the frame division, and the frame length is N,

First, judge whether each frame signal x _n can be used as an embedding area,

Then, for each frame signal x _n that can be used as an embedding area, the selection of the embedding frequency segment of the audio watermark is carried out, and the embedding frequency segment of the audio watermark is selected, and the start frequency of embedding preset according to the frequency part sensitive to the human ear is set to be FWMIN, the end frequency is FWMAX, the start embedding point freqmin1 and the embedding end point freqmax1 of a frame are calculated as follows,

freqmin1=floor((FWMIN×2.0/fs1)×N)

freqmax1=floor((FWMAX×2.0/fs1)×N)

Among them, floor is a function of rounding down;

The first time-frequency conversion module is configured to perform discrete Fourier transform on each frame signal x _n capable of embedding a watermark to obtain a frequency domain signal X _n ;

The first spread spectrum sequence generation module is used to use the key key as a random number seed to generate a binary pseudo-random spread spectrum sequence u whose length is freqmax1-freqmin1+1;

The watermark embedding module is used to embed the watermark according to the spread spectrum sequence u, the frequency domain signal X _n and the watermark bit b, and obtain the frequency domain signal after the watermark is embedded. The calculation is as follows,

|X′ _n |=|X _n |+bαu

Among them, α is a constant, which controls the embedding strength of the watermark, |X _n | and |X′ _n | represent the frequency domain amplitude before embedding the watermark and the frequency domain amplitude after embedding the watermark, and then get the embedding watermark by Euler's formula The frequency domain signal of

Among them, ∠X _n represents the phase of the frequency domain signal, X′ _n represents the frequency domain signal after embedding the watermark, and e is the mathematical natural exponent;

A time-frequency inverse transform module, which is used to transform the frequency-domain signal _X'n after the watermark is embedded into the time domain, to generate an audio file embedded with the watermark;

The watermark detection subsystem includes the following modules,

Select an appropriate area detection module for reading the audio file to be detected, and obtain the nth frame signal z _n and sampling rate fs2 after time domain framing,

First, judge whether each frame signal x _n can be used as an embedding area;

For each frame signal x _n that can be used as an embedding area, as the signal to be detected, calculate the starting point freqmin2 of the detection range and the end point freqmax2 of the frequency domain

freqmin2=floor((FWMIN×2.0/fs2)×N)

freqmax2=floor((FWMAX×2.0/fs2)×N)

The second time-frequency conversion module is used to perform discrete Fourier transform to obtain the frequency domain signal Z _n of the signal to be detected, and the corresponding frequency domain amplitude value is recorded as |Z _n |;

The second spread spectrum sequence generation module is used to use the key key as a random number seed to generate a binary pseudo-random spread spectrum sequence u whose length is freqmax2-freqmin2+1;

The correlation detection module is used to calculate the sufficient statistic _{r n of} detection according to the spread spectrum sequence u and the frequency domain amplitude value |Zn| of the signal to be detected as follows,

If the sufficient statistic r _n ≥ 0, then the detected watermark bit is b=1; otherwise, the detected watermark bit is b=0.

Moreover, in the selection of suitable area embedding module and the selection of appropriate area detection module, whether each frame signal x _n can be used as an embedding area is judged, and the implementation method is as follows,

1) Average energy of signal x _n If the size exceeds the preset corresponding threshold τ ₁ , it is a silent zone and watermarks are not allowed to be embedded;

2) If the signal x _n contains a transient signal, it is not allowed to embed the watermark.

Moreover, whether the signal x _n contains a transient signal is judged by the following method,

Assuming that a frame signal is decomposed into S blocks, the energy of the S blocks is calculated respectively, and the energy ratio rate of the block with the largest energy and the block with the smallest energy is compared with the preset corresponding threshold τ ₂ , if the rate is greater than τ ₂ , it is considered that the Frame signals contain transient signals.

The present invention proposes to filter out the transient signal to improve the accuracy of watermark detection through the ratio of the maximum energy to the minimum energy in the frame, and to improve the robustness of the watermark by embedding the watermark in the frequency band where the human ear perceives it significantly. Further, it proposes to use Averaging energy to filter out quiet areas improves perceived transparency. The technical solution of the invention has important market value.

Description of drawings

Fig. 1 is a structural block diagram of the embedded subsystem of the embodiment of the present invention.

Fig. 2 is a structural block diagram of the detection subsystem of the embodiment of the present invention.

Fig. 3 is the embedding process flowchart of the embodiment of the present invention

Fig. 4 is a flow chart of the detection process of the embodiment of the present invention.

detailed description

The technical solution of the present invention will be further described below with specific embodiments in conjunction with the accompanying drawings.

An embodiment of the present invention provides an audio watermark system based on embedding region selection, including an audio watermark embedding subsystem and a watermark detection subsystem.

Referring to Fig. 1 , the audio watermark technology embedding subsystem for embedding region selection provided by an embodiment of the present invention includes selecting a suitable region embedding module 1, a first time-frequency conversion module 2, a first spreading sequence generation module 3, and a watermark embedding module 4 and the time-frequency inverse transform module 5, each module may be realized by using software solidification technology during specific implementation.

The selection of the appropriate area embedding module 1 judges the read time-domain audio signal frame. During specific implementation, it can be judged frame by frame whether the condition for embedding the watermark can be satisfied: skip this frame if it is not satisfied, and continue the judgment of the next frame ; if satisfied, the signal is output to the first time-frequency conversion module 2, and the frequency domain signal embedded in the watermark is calculated according to the sampling rate of the time domain audio signal read and the relatively sensitive frequency range of the human ear, and can be embedded The frequency domain signal within the range is output to the watermark embedding module 4, and the maximum and minimum values of the embedding range are output to the first spread spectrum sequence generation module 3;

The first time-frequency conversion module 2 is configured to convert the read time-domain audio signal into a frequency-domain signal, and output it to the watermark embedding module 4;

The first spread spectrum sequence generation module 3 is used to generate a uniform distribution with the amplitude of the same length as the embedding range being 1 or -1 according to the random number seed and selecting the maximum value and the minimum value of the embedding range input by the appropriate area embedding module 1 random sequence, and output this random sequence to the watermark embedding module 4;

The watermark embedding module 4, for the amplitude spectrum in the frequency domain signal, generates an audio signal with watermark information in the frequency domain and outputs it to the time-frequency inverse transform module 5;

The time-frequency inverse transformation module 5 is used to convert the audio signal with watermark information in the frequency domain input by the watermark embedding module 4 into an audio signal with watermark information in the time domain, and convert the audio signal with watermark information in the time domain into an audio signal with watermark information in the time domain. The audio signal of the information generates an audio file, and an audio file with watermark information is obtained.

Referring to Fig. 2, the adaptive audio watermark detection subsystem based on phase coding provided by the embodiment of the present invention includes selecting a suitable area detection module 6, a second time-frequency conversion module 7, a second spread spectrum sequence generation module 8, and a correlation detection module 9. Each module can be realized by using software solidification technology during specific implementation.

The function of the selection suitable area detection module 6 is basically the same as that of the selection appropriate area embedding module 1, and the area that does not meet the watermark embedding conditions generally does not contain a watermark, so it can be ignored during detection: it can be judged frame by frame during specific implementation. Frames that meet the detection conditions are skipped and not detected, and the judgment of the next frame is continued; the audio signal that meets the detection conditions is output to the second time-frequency conversion module 7, and the maximum value and minimum value of the frequency detection area are also output to the second time-frequency conversion module 7. Frequency conversion module 7 and the second spreading sequence generation module 8;

The second time-frequency conversion module 7 is configured to convert the read time-domain audio signal into a frequency-domain signal, and output it to the correlation detection module 9;

The function of described second spread spectrum sequence generation module 8 and first spread spectrum sequence generation module 3 is basically the same, and the result that produces is output to correlation detection module 9;

The correlation detection module 9 is used to calculate the correlation value for the input frequency domain amplitude signal to be detected and the spread spectrum sequence input by the spread spectrum sequence generation module 9 according to the detection range, and determine the watermark according to the sign of the correlation value.

For the specific implementation of each module, refer to the corresponding steps of the method, which will not be described in detail in the present invention. The audio watermarking method based on embedding region selection provided by the embodiment of the present invention includes an embedding process and a detection process.

Referring to Fig. 3, the audio watermark embedding process based on the selected area provided by the embodiment of the present invention can be automatically carried out by means of computer software technology, and specifically includes the following steps:

Step A1, read the audio file, first divide the audio signal x in the time domain into frames, obtain the sampling rate fs1 and the nth frame time domain audio signal x _n (frame length is N) after the frame division, and for each frame signal x _n To judge whether it can be used as an embedding area, the judgment includes two aspects of judgment:

1) Judging whether the average energy of x _n exceeds the set threshold, to determine whether the current frame x _n is a silent area, if it is a silent area, it is not allowed to embed the watermark, otherwise it is not a silent area beyond the threshold, and it may be embedded. Calculate the average energy of the nth frame by the following formula

Among them, N is the frame length, that is, the number of sample points in one frame; i is the index number of sample points in one frame, and the value is between 0 and N-1; x _n ² (i) represents the time domain signal of the nth frame The energy of x _n at the i-th point in the frame; τ1 is the decision threshold _of the average energy, and those skilled in the art can preset the value during specific implementation, such as obtained according to experience; if it exceeds the threshold, then satisfy the condition 1), proceed The judgment of the following condition 2).

2) For the case of a transient signal within a frame, due to the drastic change in its frequency, it will cause a large variance. The greater the signal variance during detection, the higher the error probability of watermark detection will be. This situation should not Embed watermark. By decomposing a frame into S blocks, calculate the energy of the S blocks respectively, and compare the energy ratio rate between the block with the largest energy and the block with the smallest energy and the threshold τ ₂ , if the rate is greater than τ ₂ , it is considered that the frame signal contains Transient signals are not allowed to embed watermarks, otherwise watermarks can be embedded. During specific implementation, those skilled in the art can preset the value of S by themselves.

The specific implementation is as follows:

First, a frame signal x _{n is} divided into S blocks, then the number of sample points M in each sub-block is

M=N/S (2)

The energy E _i of each block is calculated as follows

Wherein, i represents the index number of the block in the frame, j represents the index number of the sample point in the frame, and x _n ² (j) represents the energy of the nth frame time domain signal x _n at the jth point in the frame.

Find the maximum energy E _Max and the minimum energy E _Min in the block energy

E _Max ＝MAX{E _i }, E _Min ＝MIN{E _i }, i∈[0,S-1] (4)

Among them, MAX and MIN represent the maximum value function and the minimum value function respectively.

The ratio of maximum energy to minimum energy is calculated as follows:

If rate>τ ₂ , it is considered that there is a transient signal in the signal frame x _n , and no watermark is embedded in this frame; otherwise, a watermark can be embedded. Wherein, _τ2 is a threshold value, which can be preset by those skilled in the art during specific implementation, for example, _τ2 is a detection threshold value of a transient signal, which is obtained based on experience.

Then, for each frame signal x _n that can be used as the embedding area, the selection of the embedding frequency segment of the audio watermark should be an area that is relatively obvious to the human ear. Those skilled in the art can pre-set according to the perceptual characteristics of the human ear, for example, 1000- 7000Hz. Because the signals in these areas will not be removed after filtering, audio compression and other attacks. Therefore, embedding the watermark into the perceptually obvious area will not be erased after some signal attacks and can be detected. Assuming that the preset embedding start frequency is FWMIN and the end frequency is FWMAX according to the frequency part which is sensitive to human ear perception, the corresponding start embedding point freqmin1 and embedding end point freqmax1 of one frame are calculated as follows,

freqmin1=floor((FWMIN×2.0/fs1)×N) (6)

freqmax1=floor((FWMAX×2.0/fs1)×N) (7)

Among them, floor is a function of rounding down.

According to the start embedding point freqmin1 and the embedding end point freqmax1, select the frequency domain audio signal within this range.

During specific implementation, it can be judged frame by frame, and those that do not meet the conditions are skipped, and the next frame is judged.

Step A2, perform FFT transformation (fast discrete Fourier transform) on the signal frame x _n capable of embedding the watermark into a frequency domain signal X _n .

Step A3, using the key key as a random number seed to generate a binary pseudo-random spread spectrum sequence u with a length of freqmax1-freqmin1+1.

The specific process of the embodiment in MATLAB is as follows:

First, use the key key to call the RandStream function (random seed function) to initialize the rand function (random number generation function), and then call the rand function to generate a random number, because the random number generated by the rand function is a number between 0 and 1 , it is also necessary to round these numbers into a binary pseudo-random sequence of 0 and 1, and then convert this unipolar pseudo-random sequence into a bipolar pseudo-random sequence u containing only +1 and -1.

Step A4, according to the spread spectrum sequence u, the frequency domain signal X _n and the watermark bit b, use the following formula (8) to embed the watermark to obtain the frequency domain signal after the watermark is embedded, and the calculation is realized as follows

|X′ _n |＝|X _n |+bαu (8)

Among them, α _is a constant, which controls the embedding strength of the watermark, and those skilled in the art can preset the value during specific implementation; | _Xn | Frequency-domain amplitude, and then get the frequency-domain signal after embedding the watermark through Euler's formula.

Among them, ∠X _n represents the phase of the frequency domain signal, X' _n represents the frequency domain signal after embedding the watermark, and e is the mathematical natural exponent.

Step A5, transforming the frequency domain signal X' _n embedded with the watermark into the time domain, and finally generating an audio file, that is, obtaining the audio file embedded with the watermark.

Referring to Fig. 4, the audio watermark detection process based on selected region embedding provided by the embodiment of the present invention can be automatically carried out by means of computer software technology, and specifically includes the following steps:

Step B1, read the audio file to be detected, obtain the nth frame signal z _n and sampling rate fs2 after time domain framing, and adopt the same judgment method in step A1 for each time domain signal z _n ,

That is to consider the following two conditions,

1) Average energy of signal x _n If the size exceeds the preset corresponding threshold τ ₁ , it is a silent zone and watermarks are not allowed to be embedded;

2) If the signal x _n contains a transient signal, it is not allowed to embed the watermark.

The frame signal that is not in the silent zone and does not contain transient signals can be embedded with a watermark and is to be detected.

During specific implementation, it can be judged frame by frame, and those that do not meet the conditions are skipped, and the next frame is judged.

For each frame signal x _n that can be used as an embedding area, as the signal to be detected, calculate the frequency domain start point freqmin2 and frequency domain end point freqmax2 of the detection range

freqmin2=floor((FWMIN×2.0/fs2)×N) (10)

freqmax2=floor((FWMAX×2.0/fs2)×N) (11)

In step B2, discrete Fourier transform is performed on the signal z _n satisfying the detection condition to obtain the frequency domain signal Z _n of the signal to be detected, and the corresponding frequency domain amplitude value is denoted as |Z _n |.

Step B3, use the key key to generate a binary spread spectrum sequence u (the same way as the u obtained by the embedding method above), that is, use the key key as a random number seed to generate a binary pseudo-random spread spectrum with a length of freqmax2-freqmin2+1 sequence u.

Step B4, according to the spreading sequence u and the frequency domain amplitude value | _Zn | of the signal to _be detected, by calculating the correlation value of the spreading sequence u and the frequency domain amplitude value |Zn| Statistics r _n

Among them, <·> represents the inner product calculation of the signal.

If the sufficient statistic r _n ≥ 0, then the detected watermark bit is b=1; otherwise, the detected watermark bit is b=0.

The specific embodiments described in the present invention are only to illustrate the spirit of the present invention. Those skilled in the art to which the present invention belongs can make various modifications or supplements to the described specific embodiments or adopt similar methods to replace them, but they will not deviate from the spirit of the present invention or go beyond the definition of the appended claims range.

Claims (6)

1. An audio watermarking method based on embedding region selection, characterized in that: comprising an embedding process and a detection process, the embedding process comprises the following steps, Step A1, read the audio file, obtain the sampling rate fs1 and the nth frame time-domain audio signal x _n after framing, the frame length is N, first judge whether each frame signal x _n can be used as an embedding area, Then, for each frame signal x _n that can be used as an embedding area, the selection of the embedding frequency segment of the audio watermark is carried out, and the embedding frequency segment of the audio watermark is selected, and the start frequency of embedding preset according to the frequency part sensitive to the human ear is set to be FWMIN, the end frequency is FWMAX, the start embedding point freqmin1 and the embedding end point freqmax1 of a frame are calculated as follows, freqmin1=floor((FWMIN×2.0/fs1)×N) freqmax1=floor((FWMAX×2.0/fs1)×N) Among them, floor is a function of rounding down; Step A2, performing discrete Fourier transform on each frame signal x _n capable of embedding a watermark to obtain a frequency domain signal X _n ; Step A3, using the key key as a random number seed to generate a binary pseudo-random spread spectrum sequence u with a length of freqmax1-freqmin1+1; Step A4, according to the spread spectrum sequence u, the frequency domain signal _Xn and the watermark bit b, the watermark is embedded to obtain the frequency domain signal after the watermark is embedded, and the calculation is as follows, |X′ _n |=|X _n |+bαu Among them, α is a constant, which controls the embedding strength of the watermark, |X _n | and |X′ _n | represent the frequency domain amplitude before embedding the watermark and the frequency domain amplitude after embedding the watermark, and then get the embedding watermark by Euler's formula The frequency domain signal of ${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&angle;</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}}$ Among them, ∠X _n represents the phase of the frequency domain signal, X′ _n represents the frequency domain signal after embedding the watermark, and e is the mathematical natural exponent; Step A5, transforming the frequency-domain signal X' _n embedded with the watermark into the time domain to generate an audio file embedded with the watermark; The detection process includes the following steps, Step B1, read the audio file to be detected, and obtain the nth frame signal z _n and sampling rate fs2 after time domain framing, First, judge whether each frame signal x _n can be used as an embedding area; For each frame signal x _n that can be used as an embedding area, as the signal to be detected, calculate the starting point freqmin2 of the detection range and the end point freqmax2 of the frequency domain freqmin2=floor((FWMIN×2.0/fs2)×N) freqmax2=floor((FWMAX×2.0/fs2)×N) Step B2, perform discrete Fourier transform to obtain the frequency domain signal Z _n of the signal to be detected, and record the corresponding frequency domain amplitude value as |Z _n |; Step B3, using the key key as a random number seed to generate a binary pseudo-random spread spectrum sequence u with a length of freqmax2-freqmin2+1; Step B4, according to the spread spectrum sequence u and the frequency domain amplitude value |Zn| of the signal to be detected, calculate the sufficient statistic _{r n for} detection as follows, ${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; ></span>}{<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ></span>}$ If the sufficient statistic r _n ≥ 0, then the detected watermark bit is b=1; otherwise, the detected watermark bit is b=0. 2. The audio watermarking method based on the selection of the embedding area according to claim 1, characterized in that: in step A1 and step B1, whether each frame signal xn can be used as the judgment of the embedding area _is carried out, and the implementation is as follows, 1) Average energy of signal x _n If the size exceeds the preset corresponding threshold τ ₁ , it is a silent zone and watermarks are not allowed to be embedded; 2) If the signal x _n contains a transient signal, it is not allowed to embed the watermark. 3. according to the described audio watermarking method based on embedding region selection of claim 2, it is characterized in that: whether comprise transient signal in the signal x _n , judge by the following way, Assuming that a frame signal is decomposed into S blocks, the energy of the S blocks is calculated respectively, and the energy ratio rate of the block with the largest energy and the block with the smallest energy is compared with the preset corresponding threshold τ ₂ , if the rate is greater than τ ₂ , it is considered that the Frame signals contain transient signals. 4. An audio watermarking system based on embedding region selection, characterized in that: it includes an audio watermark embedding subsystem and a watermark detection subsystem, The audio watermark embedding subsystem includes the following modules, Select an appropriate area embedding module to read the audio file, and obtain the signal x _n of the time-domain audio of the nth frame after the sampling rate fs1 and the frame division, and the frame length is N, First, judge whether each frame signal x _n can be used as an embedding area, Then, for each frame signal x _n that can be used as an embedding area, the selection of the embedding frequency segment of the audio watermark is carried out, and the embedding frequency segment of the audio watermark is selected, and the start frequency of embedding preset according to the frequency part sensitive to the human ear is set to be FWMIN, the end frequency is FWMAX, the start embedding point freqmin1 and the embedding end point freqmax1 of a frame are calculated as follows, freqmin1=floor((FWMIN×2.0/fs1)×N) freqmax1=floor((FWMAX×2.0/fs1)×N) Among them, floor is a function of rounding down; The first time-frequency conversion module is used to perform discrete Fourier transform on each frame signal x _n capable of embedding a watermark to obtain a frequency domain signal X _n ; The first spread spectrum sequence generation module is used to use the key key as a random number seed to generate a binary pseudo-random spread spectrum sequence u whose length is freqmax1-freqmin1+1; The watermark embedding module is used to embed the watermark according to the spread spectrum sequence u, the frequency domain signal X _n and the watermark bit b, and obtain the frequency domain signal after the watermark is embedded. The calculation is as follows, |X′ _n |=|X _n |+bαu Among them, α is a constant, which controls the embedding strength of the watermark, |X _n | and |X′ _n | represent the frequency domain amplitude before embedding the watermark and the frequency domain amplitude after embedding the watermark, and then get the embedding watermark by Euler's formula The frequency domain signal of ${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&angle;</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}}$ Among them, ∠X _n represents the phase of the frequency domain signal, X′ _n represents the frequency domain signal after embedding the watermark, and e is the mathematical natural exponent; A time-frequency inverse transform module, which is used to transform the frequency-domain signal _X'n after the watermark is embedded into the time domain, to generate an audio file embedded with the watermark; The watermark detection subsystem includes the following modules, Select an appropriate area detection module for reading the audio file to be detected, and obtain the nth frame signal z _n and sampling rate fs2 after time domain framing, First, judge whether each frame signal x _n can be used as an embedding area; For each frame signal x _n that can be used as an embedding area, as the signal to be detected, calculate the starting point freqmin2 of the detection range and the end point freqmax2 of the frequency domain freqmin2=floor((FWMIN×2.0/fs2)×N) freqmax2=floor((FWMAX×2.0/fs2)×N) The second time-frequency conversion module is used to perform discrete Fourier transform to obtain the frequency domain signal Z _n of the signal to be detected, and the corresponding frequency domain amplitude value is recorded as |Z _n |; The second spread spectrum sequence generation module is used to use the key key as a random number seed to generate a binary pseudo-random spread spectrum sequence u whose length is freqmax2-freqmin2+1; The correlation detection module is used to calculate the sufficient statistic _{r n of} detection according to the spread spectrum sequence u and the frequency domain amplitude value |Zn| of the signal to be detected as follows, ${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; ></span>}{<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ></span>}$ If the sufficient statistic r _n ≥ 0, then the detected watermark bit is b=1; otherwise, the detected watermark bit is b=0. 5. according to the described audio watermarking system of claim 4 based on embedding region selection, it is characterized in that: select suitable region embedding module and select suitable region detection module, carry out whether every frame signal x _n can be used as the judgment of embedding region, realize The way is as follows, 1) Average energy of signal x _n If the size exceeds the preset corresponding threshold τ ₁ , it is a silent zone and watermarks are not allowed to be embedded; 2) If the signal x _n contains a transient signal, it is not allowed to embed the watermark. 6. according to the described audio watermarking system based on embedding region selection of claim 5, it is characterized in that: whether comprise transient signal in the signal x _n , judge by the following way, Assuming that a frame signal is decomposed into S blocks, the energy of the S blocks is calculated respectively, and the energy ratio rate of the block with the largest energy and the block with the smallest energy is compared with the preset corresponding threshold τ ₂ , if the rate is greater than τ ₂ , it is considered that the Frame signals contain transient signals.