CN101271690A - Audio spread-spectrum watermark processing method for protecting audio data - Google Patents
Audio spread-spectrum watermark processing method for protecting audio data Download PDFInfo
- Publication number
- CN101271690A CN101271690A CNA200810069653XA CN200810069653A CN101271690A CN 101271690 A CN101271690 A CN 101271690A CN A200810069653X A CNA200810069653X A CN A200810069653XA CN 200810069653 A CN200810069653 A CN 200810069653A CN 101271690 A CN101271690 A CN 101271690A
- Authority
- CN
- China
- Prior art keywords
- watermark
- audio
- signal
- sequence
- noise signal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Signal Processing For Digital Recording And Reproducing (AREA)
Abstract
The invention discloses an audio spread spectrum watermark processing method for protecting the audio data, comprising two processes: a watermark embedding process and a watermark extraction process; in the embedding process, a wavelet de-noising technique is used for separating the noise of the signal and noise intensity control factor is used for reducing the noise signal intensity, and the effect of the noise of the signal on the detector performance is reduced; in order to improve the performance of the detector, the wavelet de-noising technique is also used in the water extraction process; in order to overcome the problem of excessively large correlation sliding detection calculation quantity in the corresponding detection, a synchronous positioning mechanism is designed. The algorithm is applicable to music and speech, which can effectively resist the attack of normal signal processing, more particularly, the attack of DA/AD conversion, thus being applied to the fields such as audio copyright protection, temper-proofing, certification, etc.
Description
Technical field
The present invention relates to the Information Hiding Techniques field in the multi-media information security, be specifically related to a kind of spread-spectrum watermark processing method of protecting voice data.
Background technology
The cryptographic technique that is based upon on the computational complexity theory basis is the main mode that guarantees the secret information safe transfer always, and it mainly relies on cryptographic algorithm and key length to guarantee the safety of enciphered message.Yet because the information after encrypting has randomness and not readable property, itself has just exposed the existence of ciphertext, and then suffers that the assailant to the decoding of ciphertext or to the attack of channel, causes the decoding of ciphertext or the failure of secret information transmission, causes damage.The multimedia messages concealing technology has become a big research focus of areas of information technology since the nineties in 20th century.This Information hiding is secret information to be embedded into make the ND technology of people among the digital multimedia public informations such as image, audio frequency, video; make and contain close public information and original public information in vision or acoustically be difficult to distinguish; the medium that the assailant wants in the open medium of magnanimity to find to contain secret information are extremely difficult, therefore can effectively protect the safety that needs information transmitted.What cryptographic technique was hidden is the content of secret information, and Information Hiding Techniques then is an existence of having hidden secret information, therefore has higher security.Usually, this Information Hiding Techniques is that the secret information with a kind of crying " watermark " is embedded in the multimedia digital information and stashes, in transmitting multimedia numerical information, the watermark secret information that embeds also is transmitted simultaneously, and the extraction of this secret information is confined to know the people of watermark secret information.This multimedia messages concealing technology is applied to copyright protection etc. and has positive meaning.
Digital watermark technology is as an important branch of Information Hiding Techniques application, bringing into play important effect in the copyright protection of digital product, integrated authentication and aspect such as anti-tamper.Along with internet and development of multimedia technology, unlawful activities such as piracy are spread unchecked, and it is more urgent that the demand of digital watermark technology also becomes.
Watermark handling method based on audio frequency has proposed much in recent years.Spread-spectrum watermark processing method is a relatively classic methods of a class.Traditional spread-spectrum watermark processing method (Tradi t ional Spread SpectrumEmbedding Method, brief note is TSSEM), being that the PN sequence that will represent watermark information directly is superimposed upon on the original audio signal, then is by to containing the coherent detection process that the watermark audio frequency carries out carrying out after decorrelation is handled watermark information during watermark extracting.Audio frequency by traditional spread spectrum embedding grammar embed watermark, the bit error rate that the noise effect watermark information of signal own extracts, in order to reach the lower bit error rate, therefore can only increase the embedment strength or the length of PN sequence, after yet the embedment strength of PN sequence or length increase, make can not descending by perceptual performance of watermarked audio again, the embedding capacity reduces, the ability that opposing DA/AD conversion, interchannel noise etc. are attacked.
And, a large amount of spread-spectrum watermark processing method application needs remain audio frequency watermark in simulated environment, such as preventing that the bootlegger from recording to the music of playing in the broadcasting, prevents that the assailant from distorting or the like dialog context in voice communication systems such as CDMA, GSM.These application have related to signal processings such as DA/AD conversion, voice compression coding, noise, make attacks such as the anti-DA/AD conversion of audio frequency watermark, voice compression coding become a problem that urgency is to be solved.
Because speech audio has voiceless sound, voiced sound of some characteristics that are different from other audio frequency such as music, particularly voice and quiet section alternately occur, and makes that effect was not fine when many existing robust watermarking disposal routes were used on voice.Audio-frequency watermark processing method (number of patent application: 200610036499.7) as a kind of anti-DA/AD conversion, come embed watermark in wavelet field by the energy that changes three sections wavelet coefficients, this method is applied to the music VF of non-voice, has the performance of resisting the DA/AD conversion preferably, be applied to the voice signal of low sampling rate, the bit error rate is higher.The spread-spectrum watermark processing method of time domain then is not subjected to the constraint of the characteristics of voice own, so the method for spread spectrum has wider range of application.Because DA/AD transfer process complexity, the interference noise of introducing is many, tradition spread spectrum secret information embedding grammar can only be resisted the attack that DA/AD changes by the embedment strength or the length that increase the PN sequence, and this can not satisfy well audio frequency watermark especially the voice watermark not sentience and embed the requirement of capacity, therefore need a kind of new spread-spectrum watermark processing method of design solve existing spread spectrum secret information and embed the problem that exists.
Summary of the invention
The objective of the invention is effectively to resist the problem that the DA/AD conversion is attacked at traditional spread-spectrum watermark processing method; a kind of audio spread-spectrum watermark processing method of protecting voice data is provided; this method is by based on the small echo denoising; effectively the noise of erasure signal own is handled the audio spread-spectrum watermark of correlation detector influence; be spread spectrum denoising method (Spread Spect rum Denoising Method; brief note is SSDM); not only can resist the Mp3 compression; superimposed noise; the attack that normal signals such as resampling are handled; and can effectively resist the DA/AD conversion and attack, the range of application that audio spread-spectrum watermark is handled is expanded.
The inventive method comprises:
1) at the multimedia sources end original audio information being carried out watermark embeds
A) watermark information to be embedded is carried out Error Correction of Coding with (7,4) BCH code scrambler, obtain the watermark information after the Error Correction of Coding;
B) select the PN sequence generator, obtain the PN sequence, be used to extract the key of watermark signal and the size of definite PN sequence embedment strength factor with its parameter conduct;
C) the carrier original audio is divided into acoustic signal and noise signal by wave filter, determine that according to the environment of audio frequency actual needs transmission the intensity of noise controlling elements determines the intensity of the noise signal of carrier audio frequency own, according to obtaining watermark length after the Error Correction of Coding, noise signal is carried out the branch frame, and it is identical with the length of PN sequence that each frame sampling is counted;
D) frame by frame PN sequence and the noise signal of crossing through strength control are carried out addition or additive operation according to the value of watermark, successively watermark bit is embedded in the noise signal, obtain containing the new noise signal of watermark;
E) stack of new noise signal and acoustic signal is obtained containing the audio frequency of watermark;
2) in the multimedia terminal audio-frequency information that contains watermark is carried out watermark extracting
A) sound signal that will contain watermark with wave filter is divided into acoustic signal and noise signal two parts;
B) according to the key that is used to extract watermark signal, utilize the PN sequence generator to produce the PN sequence, the watermark in the noise signal is extracted;
C) with the correlation r of slip correlation computations method calculating PN sequence and noise signal, search out the reference position that watermark first bit embeds according to correlation r;
D) from reference position, divide frame with noise signal, reorientate the sync bit of each bit watermark with synchronous reorientation method, calculate correlation r, utilize decision device to extract watermark by the polarity of judging correlation r;
E) repeating step d), extract the watermark of all process Error Corrections of Coding frame by frame;
F) utilize (7,4) BCH error correction decoder that the watermark information that extracts is decoded, obtain original watermark information.
Owing to adopted such scheme, compared with prior art, the present invention has the following advantages:
1. the bit error rate aspect of watermark extracting: adopt SSDM spread spectrum denoising method to compare with adopting TSSEM tradition spread spectrum embedding grammar, SSDM detector false code check is
TSSEM detector false code check is
When β=0, SSDM detector performance the best is disturbed if do not consider outside noise, and the SSDM bit error rate is 0, and is lower than TSSEM detector false code check
Therefore, SSDM spread spectrum denoising method has the lower bit error rate.Wherein,
Be the complementary error function, N is the length of PN sequence, and β is the noise intensity controlling elements, σ
u 2,
, σ
n 2Be respectively PN sequence u,
Be the variance of interchannel noise in noise signal after the original audio filtering and the audio signal transmission process, μ
rMathematical expectation for the correlation r that carries out obtaining after the correlation computations.
2. contain not sentience aspect of watermark audio frequency: people's ear is very responsive to the noise that signal contains, and noise intensity is big more, the easy more existence that perceives noise of people's ear.Therefore the not sentience that contains the watermark audio frequency also can be estimated with the noise variance of spatial noise, and noise variance is big more, and then noise intensity is stronger, and sentience is not poor more to contain the watermark audio frequency.The noise variance signal that contains with the audio frequency of TSSEM method embed watermark is
The noise variance signal that contains with the audio frequency of SSDM method embed watermark is
Because 0≤β≤1, so that SSDM contains the noise variance signal of watermark audio frequency is littler than TSSEM.Noise variance is more little, and people's ear not sentience is good more, and therefore, the not sentience of the moisture seal signal of SSDM is better than TSSEM.
3.SSDM the spread spectrum denoising method has been used synchronous reorientation mechanism in watermark extraction process, overcome the big shortcoming of slip correlation detector calculated amount in traditional related detecting method, reduced calculated amount effectively, this is for realizing that real-time covert communications is significant.The attack of desynchronizing can also be effectively resisted in reorientation synchronously.
4. used the error correcting encoder decoding, further reduced the bit error rate of the watermark information that embeds, made the bit error rate approach 0.
5. under the identical condition of PN sequence embedment strength, the inventive method has the lower bit error rate under the condition with very big embedding capacity, and the not sentience that contains the watermark audio frequency is better than traditional spread spectrum embedding grammar, and therefore, this method has wider range of application.Fig. 5 is that SSDM spread spectrum denoising method and TSSEM tradition spread spectrum embedding grammar embed capacity and bit error rate relation comparison diagram.In the experiment, audio frequency is that length is 10 seconds, and sampling rate is 44.1kHz, the monophonic music of 16 quantifications.M sequence embedment strength is set at 0.005, and m sequence length from 15 to 4095 does not wait (being that capacity is that 2940bit/s is to 10.8bit/s), and the noise controlling elements of SSDM are set at β=0.Use SSDM and TSSEM at the original audio embed watermark respectively, spend correlation detector then and detect watermark.Can see intuitively that by the experimental result that Fig. 5 shows during greater than 20bit/s, the detector false code check of SSDM is lower than TSSEM at the embedding capacity.Table 2 is that two kinds of methods capacity that embeds does not extract the bit error rate of watermark simultaneously.
Table 2
Experimental result by table 2 can see that for the smothing filtering detecting device, when the embedding capacity was 2940bit/s, the bit error rate of SSDM had only about 8%, and the TSSEM bit error rate is up to 34%, and the bit error rate of SSDM has descended 26%.When embedding capacity during less than 20bit/s, two kinds of method bit error rates level off to 0.
6. under the certain precondition of embedding capacity; the inventive method is under the very little situation of embedment strength; still can extract watermark information with the extremely low bit error rate; and embedment strength is little; mean that the audio frequency that contains watermark has better not sentience; therefore, have significant superiority for the copyright etc. of protection voice data, its range of application is more more extensive than TSSEM.Fig. 6 is SSDM spread spectrum denoising method and TSSEM tradition spread spectrum embedding grammar embedment strength and bit error rate relation comparison diagram.In the experiment, audio frequency is that length is 10 seconds, and sampling rate is 44.1kHz, the monophonic music of 16 quantifications.The PN sequence length is 1023 m sequence, and embedment strength from 0.0005 to 0.005 does not wait, and the noise controlling elements of SSDM are set at β=0.Use SSDM and TSSEM to the original audio embed watermark respectively, spend correlation detector then and detect watermark.Can see intuitively that by the experimental result that Fig. 6 shows less than 0.002 o'clock, the correlation detector bit error rate of SSDM was starkly lower than TSSEM at embedment strength, table 3 is bit error rates that embedment strength does not extract watermark simultaneously under two kinds of methods.
Table 3
Experimental result by table 3 can be seen, for small echo denoising correlation detector, is 0.0005 o'clock at embedment strength, and the SSDM bit error rate is 0, and the bit error rate of TSSEM is about 9%, and the detector performance of SSDM has had very big lifting.
7. especially SSDM spread spectrum denoising method has adopted the small echo denoising that the carrier original audio is divided into acoustic signal and noise signal, and in watermark extraction process, used the small echo noise-removed technology, because the small echo noise-removed technology can carry out denoising to non-stationary signals such as audio frequency very effectively, therefore the watermark correlation detector based on the small echo denoising compares medium filtering, the filtering of DCT cepstrum, traditional watermark correlation detectors such as LPC filtering and Savitzky-Golay smothing filtering have better detection performance, be under 0.0005 the situation at PN sequence embedment strength, four kinds of correlation detectors of small echo denoising correlation detector and other are compared, and bit error rate maximum can descend about 9%.Fig. 7 applies in the TSSEM tradition spread spectrum embedding grammar, and with Savitzky-Golay smothing filtering, medium filtering, LPC filtering, DCT cepstrum filtering correlation detector bit error rate comparison diagram under identical embedment strength, wherein the PN sequence length is 1023.Table 4 is to use embedment strength under the TSSEM not use different decorrelation detecting devices to extract the bit error rate of watermark simultaneously.
Table 4
Can see by the experimental result that Fig. 7 and table 4 show, at embedment strength less than 0.0025 o'clock, the detector false code check is raise successively by small echo denoising, Savitzky-Golay smothing filtering, LPC filtering, medium filtering, the filtering of DCT cepstrum, it is 0.0005 o'clock particularly at embedment strength, small echo denoising detector false code check is 9%, than the bit error rate decline about 5% of LPC filtering, medium filtering, DCT cepstrum filtering detecting device.Therefore, with respect to detecting devices such as smothing filtering, LPC filtering, medium filtering, the filtering of DCT cepstrum, use small echo denoising detector performance the best in the inventive method.
Comprehensive above-mentioned advantage, can see that the present invention has than better not sentience of TSSEM and robustness, this invention is applicable to that not only the security information of broadband audio frequency such as music hides, the security information of speech audio hide aspect more apparent its advantage, and this method can be resisted the attack of DA/AD conversion effectively, has very high practical value.
Description of drawings
Fig. 1 is the FB(flow block) that the inventive method watermark embeds;
Fig. 2 is the FB(flow block) of the inventive method watermark extracting;
Fig. 3 is the synchronous reorientation synoptic diagram of the inventive method;
Fig. 4 is original audio and the time domain waveform comparison diagram that contains the watermark audio frequency;
Fig. 5 is that SSDM spread spectrum denoising method of the present invention and TSSEM tradition spread spectrum embedding grammar embed capacity and bit error rate relation comparison diagram;
Fig. 6 is SSDM spread spectrum denoising method of the present invention and TSSEM tradition spread spectrum embedding grammar embedment strength and bit error rate relation comparison diagram;
Fig. 7 applies to small echo denoising correlation detector in the TSSEM tradition spread spectrum embedding grammar, with Savitzky-Golay smothing filtering, medium filtering, LPC filtering, DCT cepstrum filtering correlation detector bit error rate comparison diagram under identical embedment strength.
Embodiment
Referring to Fig. 1 to Fig. 3, this is the preferred embodiment that the present invention protects the audio spread-spectrum watermark processing method of voice data, comprising:
1) at the multimedia sources end original audio information is carried out watermark and embed, its detailed process is as follows:
Watermark to be embedded is that length is the random series of L, is w with vector representation, i.e. w=(w
1, w
2..., w
L), w wherein
i∈ 1,1}.
At first watermark information to be embedded is carried out Error Correction of Coding with (7,4) BCH code scrambler, obtain the watermark information after the Error Correction of Coding; Obtaining length is
Error Correction of Coding after watermark information b=(b
1, b
2..., b
7L/4), wherein might as well establish L and be 4 integral multiple, if L is not 4 integral multiple, then can behind w, add ' 1 ', benefit is 4 integral multiple.
Select the PN sequence generator, obtain the PN sequence, described PN sequence adopts the m sequence, the m sequence is the bipolar code of " 1 " or " 1 ", the key that is used to extract watermark signal with its parameter conduct, and the size of definite PN sequence embedment strength factor, PN sequence embedment strength factor value between 0.0005～0.005.
Simultaneously carrier original audio X is divided into acoustic signal B by small echo noise-removed filtering device
1With noise signal f
1, and determine that according to the environment of audio frequency actual needs transmission the intensity of noise controlling elements β determines the noise signal f of carrier audio frequency own
1Intensity.The value of described noise controlling elements β is 0≤β≤1, common intensity value between 0～0.9 of noise controlling elements β, as transmission of audio on the internet, its intensity can be between 0.5～0.9 value.According to obtaining watermark length after the Error Correction of Coding, noise signal is carried out the branch frame then, it is identical with the length of PN sequence that each frame sampling is counted.Concrete example is for according to the length that obtains watermark after the Error Correction of Coding
With noise signal f
1Be divided into
Section, every segment length is a N sampled point, the vector representation of i section noise signal is f
1i
And the PN sequence u and the noise signal f that crosses through strength control of intensity will have been determined frame by frame according to the value of watermark
1iCarry out addition or additive operation, successively watermark bit is embedded in the noise signal, be formulated as: f '
Li=β f
1i+ b
iU.Obtain containing the new noise signal f ' of watermark
1
Then with new noise signal f '
1With acoustic signal B
1Stack obtains containing the audio frequency S of watermark, both can by carrier will contain watermark only frequently S be transferred to the multimedia terminal.
2) in the multimedia terminal audio-frequency information that contains watermark is carried out watermark extracting, its detailed process is as follows:
The watermark audio frequency Y that contains that at first will receive carries out the small echo denoising with small echo noise-removed filtering device, and the sound signal that will contain watermark is divided into acoustic signal B
2With noise signal f
2Two parts, the coherent detection of watermark is at f
2In carry out.
According to the key that is used to extract watermark signal, utilize the PN sequence generator to produce the PN sequence u that has determined intensity, then to noise signal f
2In watermark extract.We are called small echo denoising detecting device whole leaching process, and concrete operations are: produce the PN sequence u that has determined intensity according to key, setting threshold a=0.6 calculates PN sequence and noise signal f with slip correlation computations method
2Correlation r, search out the reference position that watermark first bit embeds according to correlation r; When the absolute value of correlation during greater than threshold value a, stop to slide, the position of this moment as the starting point that watermark embeds, is determined that this frame is that the first bit watermark information embeds frame.
Described slip correlation computations method is calculated as follows:
Wherein,
<u，u>＝‖u‖，
In the formula: r is for calculating the value of gained, f
2Be the noise signal after the denoising of watermark audio frequency small echo of containing with vector representation, u is the PN sequence with the embedding of vector representation.
Then from reference position, with noise signal f
2Divide frame, reorientate the sync bit of each bit watermark, calculate correlation r, utilize decision device to extract watermark by the polarity of judgement correlation r with synchronous reorientation method.After containing the watermarked frame reorientation, extract watermark information according to the polarity of correlation, the watermark bit of extracting from the i section is designated as
, then extracting rule is:
。The process that described synchronous reorientation method is reorientated the sync bit of each bit watermark is: embed frame according to the first bit watermark information, the starting point that roughly the second bit watermark information is embedded frame is orientated the back sampled point that the first bit watermark information embeds frame as earlier, then from this sampled point, many respectively backward forward search Δ sampled points, calculate the correlation of 2 Δs+1 a PN sequence and noise signal, it is that the second bit watermark information embeds frame that the noise signal of the maximum correlation value that takes absolute value then correspondence is divided frame, utilizes the polarity of this correlation to extract watermark simultaneously; And then, determine that with identical process the 3rd bit watermark information embeds frame, extracts the 3rd bit watermark information according to second bit watermark information embedding frame.This process is repeated, up to the watermark that extracts all process Error Corrections of Coding frame by frame.
After extracting the watermark of all process Error Corrections of Coding, utilize (7,4) BCH error correction decoder that the watermark information that extracts is decoded at last, obtain original watermark information.
Fig. 4 is original audio and the time domain waveform comparison diagram that contains the watermark audio frequency, wherein
Fig. 4 (a) is an original audio, and original audio is that a segment length is 10 seconds, and sampling rate is 44.1kHz, the monophonic music of 16 quantifications.
Fig. 4 (b) carries out time domain waveform figure under the full embedding situation of watermark for adopting the inventive method to original audio, and parameter wherein is set to: the m sequence length is 1023, intensity 0.005, noise controlling elements β=0.Embed back signal to noise ratio snr=22.117dB.Acoustically can not differentiate original audio and the difference that contains the watermark audio frequency, utilize document " DCT territory audio frequency watermark: watermarking algorithm and not sentience estimate " (hot spring, Wang Shuxun, Nian Guijun .. electronic letters, vol, 2007,35 (9): 1702-1705.) the objective not sentience of the DMOPM of Ti Chuing (DistortionMeasurement on Psychoacoustic Model) is estimated method and is weighed the not sentience that contains the watermark audio frequency, its DMOPM value is 0.0033, this value is very little, illustrates that sentience is not fine.
Following table 1 has shown and has contained watermark music (length 10 seconds, sampling rate 44.1kHz, 16 quantifications, monophony) and voice (length 16 seconds, sampling rate 8kHz, 16 quantifications, monophony) under noise, Mp3 compression, resampling, weightization, echo, DA/AD conversion, sampling point cutting, A rule, the compression of μ rule and GSM compression, the bit error rate situation of the watermark information that is extracted under the situation that does not add BCH Error Correction of Coding and adding BCH Error Correction of Coding, parameter wherein is set to: the m sequence length is 1023, intensity 0.005.In order to make the detector false code check minimum, noise controlling elements β=0.
Attack condition in the table is: (1) DA/AD conversion: utilize realplayer player plays original audio, utilize Cool Edit Pro 2.0 to record by line input mode.(2) white Gaussian noise: average is 0, and mean square deviation is 0.01.(3) echo: delay time is 400ms, and the time delayed signal amplitude is 10% of an original signal amplitude.(4) resample: original music is resampled with 22.05kHz, resample with 44.1kHz and recover, voice then resample with 6kHz.(5) Mp3 compression: the Mp3 file with original audio boil down to position speed 128kbps and 64kbps, decompress then.(6) weightization: original audio is quantified as 8bit from 16bit.(7) sampling point cutting: every 0.05s deletes 20 sampling points continuously in original audio.(8) GSM coding: carry sound-track engraving apparatus with Windows voice are carried out the back decoding of GSM coding.(9) A rule, μ restrains compression: carry sound-track engraving apparatus with Windows and voice are carried out A restrains, μ restrains compression, decompression then.
Table 1
Can see that from experimental result the present invention is unusual robust to the signal Processing of routine, can also effectively resist the attack of DA/AD conversion.Because the bit error rate under the above-mentioned attack is lower, after adding (7,4) BCH Error Correction of Coding, the bit error rate that original watermark information is extracted is 0, can realize that the no error code of watermark information extracts.
The inventive method not only is confined to the foregoing description, except adopting small echo noise-removed filtering device, by the small echo denoising carrier original audio is divided into acoustic signal and noise signal, the sound signal that maybe will contain watermark is divided into outside acoustic signal and noise signal two parts, also can adopt the smothing filtering wave filter, by smothing filtering the carrier original audio is divided into acoustic signal and noise signal, the sound signal that maybe will contain watermark is divided into acoustic signal and noise signal two parts; Perhaps adopt the medium filtering wave filter, by medium filtering the carrier original audio is divided into acoustic signal and noise signal, the sound signal that maybe will contain watermark is divided into acoustic signal and noise signal two parts; And adopt LPC filtering, medium filtering, the filtering of DCT cepstrum etc.Comparatively speaking, its effect all is better than the effect that TSSEM tradition spread spectrum embedding grammar adopts these filtering to reach.
Claims (10)
1. audio spread-spectrum watermark processing method of protecting voice data is characterized in that described method comprises:
1) at the multimedia sources end original audio information being carried out watermark embeds
A) watermark information to be embedded is carried out Error Correction of Coding with (7,4) BCH code scrambler, obtain the watermark information after the Error Correction of Coding;
B) select the PN sequence generator, obtain the PN sequence, be used to extract the key of watermark signal and the size of definite PN sequence embedment strength factor with its parameter conduct;
C) the carrier original audio is divided into acoustic signal and noise signal by wave filter, determine that according to the environment of audio frequency actual needs transmission the intensity of noise controlling elements determines the intensity of the noise signal of carrier audio frequency own, according to obtaining watermark length after the Error Correction of Coding, noise signal is carried out the branch frame, and it is identical with the length of PN sequence that each frame sampling is counted;
D) frame by frame PN sequence and the noise signal of crossing through strength control are carried out addition or additive operation according to the value of watermark, successively watermark bit is embedded in the noise signal, obtain containing the new noise signal of watermark;
E) stack of new noise signal and acoustic signal is obtained containing the audio frequency of watermark;
2) in the multimedia terminal audio-frequency information that contains watermark is carried out watermark extracting
A) sound signal that will contain watermark with wave filter is divided into acoustic signal and noise signal two parts;
B) according to the key that is used to extract watermark signal, utilize the PN sequence generator to produce the PN sequence, the watermark in the noise signal is extracted;
C) with the correlation of slip correlation computations method calculating PN sequence and noise signal, search out the reference position that watermark first bit embeds according to correlation;
D) from reference position, divide frame with noise signal, reorientate the sync bit of each bit watermark with synchronous reorientation method, calculate correlation, utilize decision device to extract watermark by the polarity of judging correlation;
E) repeating step d), extract the watermark of all process Error Corrections of Coding frame by frame;
F) utilize (7,4) BCH error correction decoder that the watermark information that extracts is decoded, obtain original watermark information.
2. the audio spread-spectrum watermark processing method of protection voice data according to claim 1 is characterized in that: described PN sequence adopts the m sequence, and the m sequence is the bipolar code of " 1 " or " 1 ".
3. the audio spread-spectrum watermark processing method of protection voice data according to claim 1 is characterized in that: described PN sequence embedment strength factor value between 0.0005～0.005.
4. the audio spread-spectrum watermark processing method of protection voice data according to claim 1 is characterized in that: the intensity of described noise controlling elements value between 0～0.9.
5. the audio spread-spectrum watermark processing method of protection voice data according to claim 1 is characterized in that: described slip correlation computations method is calculated as follows:
Wherein,
<u，u>＝‖u‖，
In the formula: r is for calculating the value of gained, f
2For with vector representation contain the filtered noise signal of watermark audio frequency, u is the PN sequence with the embedding of vector representation.
6. the audio spread-spectrum watermark processing method of protection voice data according to claim 1; it is characterized in that the reference position that described searching watermark first bit embeds is: produce the PN sequence of having determined intensity according to the key that is used to extract watermark signal; setting threshold a=0.6; utilize the PN sequence and contain the filtered noise signal of watermark audio frequency and make the slip correlation computations; when the absolute value of correlation during greater than threshold value a; stop to slide; position at this moment as the reference position that watermark embeds, is determined that this frame is that the first bit watermark information embeds frame.
7. the audio spread-spectrum watermark processing method of protection voice data according to claim 1, it is characterized in that the process that described synchronous reorientation method is reorientated the sync bit of each bit watermark is: embed frame according to the first bit watermark information, the starting point that roughly the second bit watermark information is embedded frame is orientated the back sampled point that the first bit watermark information embeds frame as earlier, then from this sampled point, many respectively backward forward search Δ sampled points, calculate the correlation of 2 Δs+1 a PN sequence and noise signal, it is that the second bit watermark information embeds frame that the noise signal of the maximum correlation value that takes absolute value then correspondence is divided frame, utilizes the polarity of this correlation to extract watermark simultaneously; And then, determine that with identical process the 3rd bit watermark information embeds frame, extracts the 3rd bit watermark information according to second bit watermark information embedding frame.This process is repeated, up to extracting all watermarks.
8. according to the audio spread-spectrum watermark processing method of the arbitrary described protection voice data of claim 1～7; it is characterized in that: described wave filter adopts small echo noise-removed filtering device; by the small echo denoising carrier original audio is divided into acoustic signal and noise signal, the sound signal that maybe will contain watermark is divided into acoustic signal and noise signal two parts.
9. according to the audio spread-spectrum watermark processing method of the arbitrary described protection voice data of claim 1～7; it is characterized in that: described wave filter adopts the smothing filtering wave filter; by smothing filtering the carrier original audio is divided into acoustic signal and noise signal, the sound signal that maybe will contain watermark is divided into acoustic signal and noise signal two parts.
10. according to the audio spread-spectrum watermark processing method of the arbitrary described protection voice data of claim 1～7; it is characterized in that: described wave filter adopts the medium filtering wave filter; by medium filtering the carrier original audio is divided into acoustic signal and noise signal, the sound signal that maybe will contain watermark is divided into acoustic signal and noise signal two parts.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810069653XA CN101271690B (en) | 2008-05-09 | 2008-05-09 | Audio spread-spectrum watermark processing method for protecting audio data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810069653XA CN101271690B (en) | 2008-05-09 | 2008-05-09 | Audio spread-spectrum watermark processing method for protecting audio data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101271690A true CN101271690A (en) | 2008-09-24 |
| CN101271690B CN101271690B (en) | 2010-12-22 |
Family
ID=40005606
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200810069653XA Expired - Fee Related CN101271690B (en) | 2008-05-09 | 2008-05-09 | Audio spread-spectrum watermark processing method for protecting audio data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101271690B (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102208187A (en) * | 2010-03-30 | 2011-10-05 | 尼尔森（美国）有限公司 | Methods and apparatus for audio watermarking a substantially silent media content presentation |
| CN102708869A (en) * | 2012-05-30 | 2012-10-03 | 武汉大学 | Robust digital audio watermarking method |
| CN102890933A (en) * | 2012-09-14 | 2013-01-23 | 瑞声声学科技(深圳)有限公司 | Audio signal collecting device, audio signal transmitting method and system |
| CN103299366A (en) * | 2011-02-07 | 2013-09-11 | 高通股份有限公司 | Devices for encoding and detecting a watermarked signal |
| CN103650039A (en) * | 2011-07-08 | 2014-03-19 | 汤姆逊许可公司 | Method and apparatus for quantisation index modulation for watermarking an input signal |
| CN104658542A (en) * | 2015-03-16 | 2015-05-27 | 武汉大学 | Additive spread spectrum audio watermarking embedding method, additive spread spectrum audio watermarking detection method and additive spread spectrum audio watermarking embedding system based on orthogonality |
| CN105244033A (en) * | 2014-07-09 | 2016-01-13 | 意法半导体亚太私人有限公司 | System and method for digital watermarking |
| CN105337942A (en) * | 2014-08-07 | 2016-02-17 | 联想(北京)有限公司 | Information processing method and electronic equipment |
| CN105721099A (en) * | 2016-01-19 | 2016-06-29 | 努比亚技术有限公司 | Terminal and near-field communication method |
| CN106209297A (en) * | 2016-07-05 | 2016-12-07 | 华中科技大学 | Radio frequency watermark embedding based on the modulation of QPSK signal and extracting method and system |
| CN106254345A (en) * | 2016-08-04 | 2016-12-21 | 安徽大学 | A kind of mobile terminal sound communication integrity certification devices and methods therefor |
| CN106415713A (en) * | 2014-04-02 | 2017-02-15 | 彼得·格雷厄姆·克雷文 | Transparent lossless audio watermarking |
| US9767822B2 (en) | 2011-02-07 | 2017-09-19 | Qualcomm Incorporated | Devices for encoding and decoding a watermarked signal |
| CN108475510A (en) * | 2015-12-23 | 2018-08-31 | 马尔科姆·罗 | Transparent lossless audio frequency watermark enhancing |
| CN109816599A (en) * | 2018-12-07 | 2019-05-28 | 西安电子科技大学 | A kind of image band noise suppressing method based on wavelet decomposition convolutional neural networks |
| CN111429901A (en) * | 2020-03-16 | 2020-07-17 | 云知声智能科技股份有限公司 | IoT chip-oriented multi-stage voice intelligent awakening method and system |
| CN113178199A (en) * | 2021-06-29 | 2021-07-27 | 中国科学院自动化研究所 | Digital audio tampering evidence obtaining method based on phase deviation detection |
| CN113362835A (en) * | 2020-03-05 | 2021-09-07 | 杭州网易云音乐科技有限公司 | Audio watermark processing method and device, electronic equipment and storage medium |
| CN115662444A (en) * | 2022-12-14 | 2023-01-31 | 北京惠朗时代科技有限公司 | Electronic seal voice interactive application method and system based on artificial intelligence |
| CN118138852A (en) * | 2024-05-08 | 2024-06-04 | 中国人民解放军国防科技大学 | Audio digital watermark embedding method and device |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104125052A (en) * | 2014-07-24 | 2014-10-29 | 清华大学 | Parallel timing synchronization system and method |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102004021404B4 (en) * | 2004-04-30 | 2007-05-10 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Watermark embedding |
| CN1848829B (en) * | 2005-04-14 | 2010-06-16 | 北京中科信利技术有限公司 | Method for automatic synchronizing of audio-frequency watermark |
-
2008
- 2008-05-09 CN CN200810069653XA patent/CN101271690B/en not_active Expired - Fee Related
Cited By (34)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102208187A (en) * | 2010-03-30 | 2011-10-05 | 尼尔森（美国）有限公司 | Methods and apparatus for audio watermarking a substantially silent media content presentation |
| CN103299366B (en) * | 2011-02-07 | 2015-06-10 | 高通股份有限公司 | Devices for encoding and detecting a watermarked signal |
| CN103299366A (en) * | 2011-02-07 | 2013-09-11 | 高通股份有限公司 | Devices for encoding and detecting a watermarked signal |
| US9767822B2 (en) | 2011-02-07 | 2017-09-19 | Qualcomm Incorporated | Devices for encoding and decoding a watermarked signal |
| US9767823B2 (en) | 2011-02-07 | 2017-09-19 | Qualcomm Incorporated | Devices for encoding and detecting a watermarked signal |
| CN103650039A (en) * | 2011-07-08 | 2014-03-19 | 汤姆逊许可公司 | Method and apparatus for quantisation index modulation for watermarking an input signal |
| US10019997B2 (en) | 2011-07-08 | 2018-07-10 | Thomson Licensing | Method and apparatus for quantisation index modulation for watermarking an input signal |
| CN103650039B (en) * | 2011-07-08 | 2016-08-03 | 汤姆逊许可公司 | For the method and apparatus that the quantification index of input signal watermarking is modulated |
| CN102708869A (en) * | 2012-05-30 | 2012-10-03 | 武汉大学 | Robust digital audio watermarking method |
| CN102890933A (en) * | 2012-09-14 | 2013-01-23 | 瑞声声学科技(深圳)有限公司 | Audio signal collecting device, audio signal transmitting method and system |
| CN106415713B (en) * | 2014-04-02 | 2021-08-17 | Mqa 有限公司 | Method for transparent lossless watermarking of audio |
| CN106415713A (en) * | 2014-04-02 | 2017-02-15 | 彼得·格雷厄姆·克雷文 | Transparent lossless audio watermarking |
| CN105244033A (en) * | 2014-07-09 | 2016-01-13 | 意法半导体亚太私人有限公司 | System and method for digital watermarking |
| CN105337942A (en) * | 2014-08-07 | 2016-02-17 | 联想(北京)有限公司 | Information processing method and electronic equipment |
| CN104658542B (en) * | 2015-03-16 | 2018-01-12 | 武汉大学 | Based on orthogonal additivity spread spectrum audio frequency watermark embedding grammar, detection method and system |
| CN104658542A (en) * | 2015-03-16 | 2015-05-27 | 武汉大学 | Additive spread spectrum audio watermarking embedding method, additive spread spectrum audio watermarking detection method and additive spread spectrum audio watermarking embedding system based on orthogonality |
| CN108475510A (en) * | 2015-12-23 | 2018-08-31 | 马尔科姆·罗 | Transparent lossless audio frequency watermark enhancing |
| CN105721099A (en) * | 2016-01-19 | 2016-06-29 | 努比亚技术有限公司 | Terminal and near-field communication method |
| CN106209297A (en) * | 2016-07-05 | 2016-12-07 | 华中科技大学 | Radio frequency watermark embedding based on the modulation of QPSK signal and extracting method and system |
| CN106209297B (en) * | 2016-07-05 | 2018-11-02 | 华中科技大学 | Radio frequency watermark insertion and extracting method based on QPSK signal modulations and system |
| CN106254345A (en) * | 2016-08-04 | 2016-12-21 | 安徽大学 | A kind of mobile terminal sound communication integrity certification devices and methods therefor |
| CN106254345B (en) * | 2016-08-04 | 2019-06-04 | 安徽大学 | A kind of mobile terminal sound communication integrity authentication device and its method |
| CN109816599B (en) * | 2018-12-07 | 2020-12-29 | 西安电子科技大学 | Image banding noise suppression method based on wavelet decomposition convolution neural network |
| CN109816599A (en) * | 2018-12-07 | 2019-05-28 | 西安电子科技大学 | A kind of image band noise suppressing method based on wavelet decomposition convolutional neural networks |
| CN113362835A (en) * | 2020-03-05 | 2021-09-07 | 杭州网易云音乐科技有限公司 | Audio watermark processing method and device, electronic equipment and storage medium |
| CN113362835B (en) * | 2020-03-05 | 2024-06-07 | 杭州网易云音乐科技有限公司 | Audio watermarking method, device, electronic equipment and storage medium |
| CN111429901A (en) * | 2020-03-16 | 2020-07-17 | 云知声智能科技股份有限公司 | IoT chip-oriented multi-stage voice intelligent awakening method and system |
| CN111429901B (en) * | 2020-03-16 | 2023-03-21 | 云知声智能科技股份有限公司 | IoT chip-oriented multi-stage voice intelligent awakening method and system |
| CN113178199A (en) * | 2021-06-29 | 2021-07-27 | 中国科学院自动化研究所 | Digital audio tampering evidence obtaining method based on phase deviation detection |
| CN113178199B (en) * | 2021-06-29 | 2021-08-31 | 中国科学院自动化研究所 | Digital audio tampering evidence obtaining method based on phase deviation detection |
| US11521629B1 (en) | 2021-06-29 | 2022-12-06 | Institute Of Automation, Chinese Academy Of Sciences | Method for obtaining digital audio tampering evidence based on phase deviation detection |
| CN115662444A (en) * | 2022-12-14 | 2023-01-31 | 北京惠朗时代科技有限公司 | Electronic seal voice interactive application method and system based on artificial intelligence |
| CN118138852A (en) * | 2024-05-08 | 2024-06-04 | 中国人民解放军国防科技大学 | Audio digital watermark embedding method and device |
| CN118138852B (en) * | 2024-05-08 | 2024-07-09 | 中国人民解放军国防科技大学 | Audio digital watermark embedding method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101271690B (en) | 2010-12-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101271690A (en) | Audio spread-spectrum watermark processing method for protecting audio data | |
| Kirovski et al. | Spread-spectrum watermarking of audio signals | |
| Djebbar et al. | A view on latest audio steganography techniques | |
| Dutta et al. | Data hiding in audio signal: A review | |
| JP2001282265A (en) | Speech data hiding method and device by computer | |
| US20030176934A1 (en) | Method and apparatus for embedding data in audio signals | |
| Gopalan | Audio steganography by cepstrum modification | |
| CN102074238A (en) | Linear interference cancellation-based speech secrete communication method | |
| Dhar et al. | Digital watermarking scheme based on fast Fourier transformation for audio copyright protection | |
| Park et al. | Speech authentication system using digital watermarking and pattern recovery | |
| Attari et al. | Robust and transparent audio watermarking based on spread spectrum in wavelet domain | |
| CN101350198B (en) | Method for compressing watermark using voice based on bone conduction | |
| Zhang et al. | Robust and transparent audio watermarking based on improved spread spectrum and psychoacoustic masking | |
| Cui et al. | The application of binary image in digital audio watermarking | |
| Wei et al. | Audio watermarking using time-frequency compression expansion | |
| Roy et al. | An efficient and blind audio watermarking technique in DCT domain | |
| El-Khamy et al. | Chaos-based image hiding scheme between silent intervals of high quality audio signals using feature extraction and image bits spreading | |
| Patel et al. | Secure transmission of password using speech watermarking | |
| Wu et al. | Comparison of two speech content authentication approaches | |
| Trivedi et al. | An algorithmic digital audio watermarking in perceptual domain using direct sequence spread spectrum | |
| Xu et al. | Content-based digital watermarking for compressed audio | |
| Yan et al. | Improved echo hiding using power cepstrum and simulated annealing based synchronization technique | |
| Hu et al. | A perceptually adaptive and retrievable QIM scheme for efficient blind audio watermarking | |
| Esmaili et al. | A novel spread spectrum audio watermarking scheme based on time-frequency characteristics | |
| Li et al. | Spread-spectrum audio watermark robust against pitch-scale modification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20101222 Termination date: 20120509 |