CN102222504A - Digital audio multilayer watermark implanting and extracting method - Google Patents

Abstract

The invention discloses a digital audio multilayer watermark implanting method, comprising the following steps: a. determining an ith frequency band inserted into an ith layer of watermark in an original audio signal; b. judging whether i-1 is zero when the ith layer of watermark is inserted, if not, the step c is carried out, if so, whether an (i-1)th layer watermark is implanted in an (i-1)th frequency band is judged continuously, and if so, the step c is carried out; c. the ith layer of watermark is inserted into the ith frequency band in the original audio signal, wherein i is more than or equal to 1 and is less than or equal to 4, and i is a positive integer. The invention also discloses a digital audio multilayer watermark extracting method. In the invention, by utilizing the Fourier transform of a quantizing audio signal, the audio signal is converted to a frequency domain from a time domain to implant or extract a watermark code, and is expanded to a specific multi-layer frequency spectrum, and high reliability is realized.

Description

Watermark implantation of DAB multilayer and extracting method

Technical field

The present invention relates to Digital Signal Processing, especially relate to a kind of method of implanting and extracting at the multilayer digital watermark information of original audio digital signal.

Background technology

The mark watermark is meant in the medium of confidential information embedding as video, image, audio frequency and text that will be called " watermark ".The watermark information that embeds can only limit to be extracted by the people who understands this information.Medium with watermark are identical with generic media In the view of the general user.

Especially owing to compare with similar medium, wherein by electric wave or communication network, access, transmission, editor and keeping easily and do not produce the advantage of data degradation during DATA DISTRIBUTION, Digital Media produces the problem of new copyright protection.Digital watermarking comes on the scene as the method for protecting copyright.

Digital watermarking not only is used for insertion information and distinguishes the owner with the protection copyright, also be used for inserting anti-pirate, distribute and confirm the control information of (distribution confirmation), broadcasting monitoring etc. or be used for information as the demonstration time control information, (lip-sync), content information and the lyrics are inserted into real-time media such as audio frequency, video etc. and transmit the information of inserting synchronously.

According to range of application, can be divided into still image watermark, video image watermark and audio frequency watermark to digital watermarking usually.In recent years, research to still image watermark and video image watermark is a lot, and it is fewer for the digital audio frequency watermark Study on Technology, this is owing to compare with video with image, much less is wanted in counting that sound signal was sampled in each time interval, and means that implantable information is than visual media much less in the sound signal.In addition, human auditory system (HAS) is much sensitiveer than human visual system (HVS), can not aesthesia implementing than difficulty visually acoustically.

In water mark method, the sound signal water mark method has multiple example, and (least significant bit (LSB): coding method Least Significant Bit), echo are hidden (echo hiding) method, reached spread-spectrum (spread spectrum) communication means etc. as LSB.

In the LSB coding method, the least significant bit (LSB) of the audio samples of quantification is deformed to insert the information of expectation.The LSB coding method has utilized the distortion of sound signal least significant bit (LSB) to influence this feature of tonequality hardly.The advantage of LSB coding method is, insert and detection is easy to carry out and less distortion of tonequality, but its shortcoming is, and is vulnerable as diminishing compression or filtering for signal Processing.

In addition, in the echo hidden method, a unheard echo is inserted in the sound signal.That is to say that the echo hidden method is according to the scale-of-two watermark information that will be inserted into, the echo that will have the different time delay is inserted and is encoded in the sound signal of segmenting with predetermined space.In decode procedure, postpone by the echo time in the duration of detecting each segmentation, binary message is decoded.Like this, the signal of insertion is not a noise, but has the sound signal itself with the original signal identical characteristics.Therefore, even the signal that inserts is heard, the signal of insertion can not be identified as distorted signal yet.Insert signal and also be considered to provide better tone color.Therefore, the echo hidden method is applicable to the high quality audio watermark, but its shortcoming is, because detect by cepstrum (Cepstrum) operation, the operation amount of decoding is very high, and if lack synchronously in the duration of time domain segmentation, just can't decode.

In addition, the spread spectrum communication method is a kind of typical water mark method, and it is popular and be used for the research of audio frequency watermark at most in video watermark.In the spread spectrum communication method, sound signal is transformed to frequency by discrete Fourier transform (DFT) (DFT), and thereafter the scale-of-two watermark information by frequency expansion to PN (pseudo noise) sequence so that extend information is inserted in the sound signal of frequency transformation.Utilize the high auto-associating characteristic of PN sequence, the watermark of insertion can be passed through correlator (correlator) and detect, and has the characteristic of jamproof robustness and good encryption.Otherwise its shortcoming of spread spectrum communication method is: sound quality descends, and the operation amount that inserts and detect is very high, and if can not be done being inserted with high-power watermark compressed encoding when strengthening its robustness.

Like this, summarize traditional audio frequency watermark, traditional audio frequency watermark shortcoming is: because usually original signal is compressed and decodes before watermark information therefore be inserted in the original signal its implementation complexity, more operation amount and the original signal easy deformation of needs when it compresses.

Summary of the invention

The present invention proposes a kind of digital audio multilayer digital watermarking implantation and extracting method, to solve the technical matters that the current audio watermark is inserted and extracted complicated operation, is difficult to realize.

The present invention adopts following technical scheme to realize: a kind of DAB multilayer watermark method for implantation, and it comprises step:

The i frequency band of a, definite insertion i layer watermark in former sound signal;

B, when inserting the watermark of i layer, judge whether i-1 is 0, if be not 0, change step c over to, otherwise continuation judges whether implanted the watermark of i-1 layer in the i-1 frequency band, if then change step c over to;

C, insertion i layer watermark in the i frequency band of former sound signal;

Wherein, 4 〉=i 〉=1, and i is a positive integer.

In a preferred embodiment, step c comprises:

C1, determine the watermark code of i layer watermark;

C2, read the audio signal segment of schedule time length successively, in each audio signal segment, implant watermark code respectively, the audio signal segment of implanting watermark code is written to a new sound signal.

In a preferred embodiment, step c2 comprises:

C21, implantation moment of calculating each correspondence on the time shaft of former sound signal in binary watermark code;

C22, the former sound signal that will implant are constantly done Fourier transform, and time domain is transformed to frequency field, calculate the new amplitude of i frequency band after the watermark code of implanting the watermark of i layer in frequency field.

In a preferred embodiment, the calculation procedure in the new amplitude of i frequency band described in the step c22 comprises:

Each implants constantly the upper limiting frequency f of i frequency band in the former sound signal watermark code of c221, the watermark of calculating i layer _iWith lower frequency limit g _iCorresponding respectively amplitude A _Fi, A _Gi:

$A_{\mathrm{fi}}=\sqrt{{\left[\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}\cos \left(\frac{{2\mathrm{\πf}}_i}{F_s}\right)x_n\right]}^2+{\left[\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}\sin \left(\frac{2{\mathrm{\πf}}_i}{F_s}\right)x_n\right]}^2},$

$A_{\mathrm{gi}}=\sqrt{{\left[\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}\cos \left(\frac{{2\mathrm{\πg}}_i}{F_s}\right)x_n\right]}^2+{\left[\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}\sin \left(\frac{2{\mathrm{\πg}}_i}{F_s}\right)x_n\right]}^2},$

If the watermark code binary digit that c222 implants is 1, and

The upper limiting frequency f of i frequency band in the time of then should implanting constantly _iWith lower frequency limit g _iThe corresponding respectively following calculating of new amplitude:

Otherwise the upper limiting frequency f of i frequency band _iWith lower frequency limit g _iDui Ying new amplitude equals the amplitude A that calculates among the step c221 respectively _FiAnd A _Gi

If the watermark code binary digit that c223 implants is 0, and

The upper limiting frequency f of i frequency band in the time of then should implanting constantly _iWith lower frequency limit g _iThe corresponding respectively following calculating of new amplitude:

Otherwise the upper limiting frequency f of i frequency band _iWith lower frequency limit g _iDui Ying new amplitude equals the amplitude A that calculates among the step c221 respectively _FiAnd A _Gi

Wherein, N represents the sampled point number of former sound signal; x _nBe former sampled audio signal value sequence, F _sBe sample frequency; γ _sBe user's controlled variable.

In a preferred embodiment, 1＜γ _s＜10.

In a preferred embodiment, describedly judging whether implanted the watermark of i-1 layer in the i-1 frequency band, is by the i-1 frequency band being carried out the watermark extracting operation, if find watermark, promptly thinks and has implanted the watermark of i-1 layer in the i-1 frequency band.

In a preferred embodiment, watermark code be one between 16 to 32 binary numeral.

In a preferred embodiment, the 1st frequency band is 500Hz-600Hz; The 2nd frequency band is 550Hz-650Hz; The 3rd frequency band is 700Hz-800Hz; The 4th frequency band is 750Hz-850Hz.

In addition, the present invention also discloses a kind of DAB multilayer watermark extracting method, and it comprises step:

To the frequency band of user's appointment, or 4 frequency bands from giving tacit consent to, carry out the watermark extracting operation;

Determine the time information of watermark code in sound signal in the i frequency band by two frequency detectors;

Convert sound signal to frequency field by Fourier transform, in frequency field, calculate the watermark code place amplitude of i frequency band constantly;

By i frequency band upper limiting frequency f _iCorresponding amplitude

With lower frequency limit g _iCorresponding amplitude

Amplitude ratio:

Obtain the binary sequence B of watermark code _i, wherein $B_i=\left\{\begin{array}{c}1,({\mathrm{\&Delta;}}_i\&GreaterEqual;1)\\ 0,({\mathrm{\&Delta;}}_i<1)\end{array}\right..$

In a preferred embodiment, the amplitude of i frequency band upper limiting frequency correspondence

The amplitude corresponding with lower frequency limit

Be respectively:

${\hat{A}}_{\mathrm{fi}}=\sqrt{{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\cos \left(\frac{{2\mathrm{\&pi;f}}_i}{F_s}\right)y_n\right]}^2+{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\sin \left(\frac{2{\mathrm{\&pi;f}}_i}{F_s}\right)y_n\right]}^2};$

${\hat{A}}_{\mathrm{gi}}=\sqrt{{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\cos \left(\frac{{2\mathrm{\&pi;g}}_i}{F_s}\right)y_n\right]}^2+{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\sin \left(\frac{2{\mathrm{\&pi;g}}_i}{F_s}\right)y_n\right]}^2};$

Wherein, y _nThe sound signal of expression band watermark, N represents the sampled point number of former sound signal, F _sBe sample frequency.

Compared with prior art, the present invention has following beneficial effect:

The present invention utilizes the Fourier transform that quantizes sound signal; sound signal is transformed into frequency field from time domain implants or extract watermark code; and extend on specific 3 layers or the 4 layers of frequency spectrum; thereby also avoided existing digital watermark all only to embed one deck watermark basically, and can't form a difficult problem the protection of music knowledge property right industrial chain.

In addition, the present invention only carries out very small change to former sound signal, and the digital watermarking that this means adding is ND concerning user's the sense of hearing and vision.Unless original signal is carried out huge destructive destruction, otherwise the watermark that adds can not be worn away or be revised also.Watermark simultaneously has high reliability, can resist various format conversion, noise, and malice is distorted and is attacked.

Description of drawings

Fig. 1 is that the present invention realizes the schematic flow sheet that the watermark of DAB multilayer is implanted;

Fig. 2 is the idiographic flow synoptic diagram of step S14 among Fig. 1.

Embodiment

The present invention proposes an implementation method that embeds and extract based on the digital watermarking of voice data.Characteristic at sound signal, utilize the variation relation of energy between a plurality of adjacent spectra of sound signal, embed digital watermark and do not influence the quality of audio frequency, bear the attack of compression, low-pass filtering, standardization, D/A conversion and signal cutting respectively, still can keep the integrality of digital watermarking.When voice data was authenticated, this digital watermarking detecting and the technology of fetching did not need original audio signal, just digital watermark data can be taken out fast.Application error is corrected sign indicating number and synchronous code method for searching in digital watermarking simultaneously, make when music is attacked, can improve the return rate of watermark data greatly, even after audio frequency is transformed form, for example from the wav format conversion to the mp3 form, digital watermarking can not lost yet.

The present invention utilizes fast horizontal auto adapted filtering (the Fast Transversal Filter that quantizes sound signal; FTF) coefficient; and extend on the specific multilayer frequency spectrum of sound signal; thereby also avoided existing digital watermark all only to embed one watermark basically, and can't form a difficult problem the protection of music knowledge property right industrial chain.

Because the quantity of information that can hold in the audio file is limited, if watermark information is too many, can seriously undermine the quality of original audio, damages.Therefore, the number of plies of the watermark code of generally inserting in audio file can not be too big.In a preferred embodiment of the invention, in audio file, insert four layers of watermark code, supply chain relationship during the three first layers hierarchical relationship of watermark code can be used for reflecting reality,---publisher (the Distributor)---user (User) that is copyright side (Owner), perhaps also can be used for representing parallel issuer of 2 or 3 families, the 4th layer can give over to standby.

As shown in Figure 1, the DAB multilayer watermark method for implantation of the present invention's proposition comprises performing step:

Step S11: the frequency band of determining each layer watermark insertion sound signal (former sound signal).

The 1st frequency band (f of former sound signal is inserted in the 1st layer of watermark ₁, g ₁); The 2nd frequency band (f of former sound signal is inserted in the 2nd layer of watermark ₂, g ₂); The 3rd frequency band (f of former sound signal is inserted in the 3rd layer of watermark ₃, g ₃); The 4th frequency band (f of former sound signal is inserted in the 4th layer of watermark ₄, g ₄).Wherein, the 1st frequency band (f ₁, g ₁), the 2nd frequency band (f ₂, g ₂), the 3rd frequency band (f ₃, g ₃), the 4th frequency band (f ₄, g ₄) be the frequency band in 100Hz to the 4000Hz frequency range.

In these 4 frequency bands, f ₁, g ₁, f ₂, g ₂, f ₃, g ₃, f ₄And g ₄All representing frequency, default value is: f ₁=500Hz, g ₁=600Hz, f ₂=550Hz, g ₂=650Hz, f ₃=700Hz, g ₃=800Hz, f ₄=750Hz, g ₄=850Hz.

Because in sound signal, voice and various music are a kind of random signals, its waveform is very complicated, and the frequency range of sub-audible sound generally can reach 20Hz-20KHz; Wherein the spectral range of language is about about 150Hz-4KHz; And the spectral range of various music can reach about 40Hz-18KHz; The energy distribution of its average frequency spectrum is: bass and middle bass part maximum, and alt partly takes second place, high pitch part minimum (in being about, bass part energy 1/10); The energy 90% of voice mainly concentrates on the 200Hz-3.5KHz frequency range.The big 10-15dB (even higher) of its mean value of peakedness ratio of these audible sound random signal amplitudes.According to such principle, selecteed 4 frequency bands have been concentrated most of energy of original audio signal.

Determine that like this frequency band that inserts watermark has two aspect benefits: the first, it guarantees have the duration of abundant amount to implant watermark code, is used for implementing the noise protection; The second, in such frequency band was selected, when former sound signal was compressed, the waveform of former sound signal can guarantee safety basically, for example, and when the wav File Compress is converted to the mp3 form.

Step S12: when inserting i layer (i 〉=1, and i is a positive integer) watermark, judge whether i-1 is 0.

The implantation logical order of watermark code is followed successively by the 1st layer of watermark, the 2nd layer of watermark, the 3rd layer of watermark and the 4th layer of watermark.If do not implant the 1st layer of watermark, can't implant the 2nd layer of watermark, in like manner, do not implant the 2nd layer of watermark, can't implant the 3rd layer of watermark.Guaranteeing the implantation order of watermark, also is a kind of protection to industry distribution chain rule.Certainly, the implantation time of this every layer watermark is not limit, and after implanting the 1st layer of watermark, implants the 2nd layer of watermark again after 1 year.The present invention can detect whether implanted preceding one deck watermark (except the ground floor watermark) earlier before each the implantation.Watermark just can't be erased watermark from audio frequency after implanting again.

Sequencing between inserting based on which floor watermark when inserting i layer (i 〉=1, and i is a positive integer) watermark, needs judge whether its last layer (i-1 layer) has implanted watermark earlier; Wherein, during as if the 1st layer of watermark of implantation, there is not the last layer watermark in it, thereby need not to judge whether implanted the last layer watermark in this audio file.Therefore, this step S12 is by judging whether i-1 is 0, if equal 0, illustrating current is to implant the 1st layer of watermark, otherwise for implanting the 2nd, 3 or 4 layer of watermark.

If i-1=0 then changes step S14 over to, otherwise changes step S13 over to.

Step S13: judge whether the i-1 layer has implanted watermark, has implanted watermark as the i-1 layer, then changes step S14 over to, otherwise can't implant the watermark of i layer.

4 layers of watermark corresponding respectively above-mentioned 4 frequency bands, if the user specifies and need write i layer (i is 2,3 or 4) watermark, then at first can be to the i-1 frequency band (f of i-1 layer watermark in correspondence _I-1, g _I-1) carry out the watermark extracting operation, if find watermark, promptly think and implanted the last layer watermark.If do not find frequency band (f _I-1, g _I-1) in comprise watermark, then think and do not implant watermark, just point out the user to need to implant earlier the watermark of i-1 layer

Step S14: at the i frequency band (f of former sound signal _i, g _i) watermark of insertion i layer.

In conjunction with shown in Figure 2, specifically, this step S14 comprises:

Step S141: set and insert former sound signal i frequency band (f _i, g _i) in the numerical value (or watermark code) of i layer watermark.

The watermark implantation process be in original digital audio and video signals, implant one default and have the watermark code of uniqueness, the numeral of watermark is that a length is m _bThe binary number of position, satisfy:

16≤m _b≤32。

In addition, the numeral of watermark can be by consumer premise justice code table or the code table that uses preparation in advance to distribute.

Can implant the separate watermark code of multilayer in a sound signal (such as a first song), under the default situation, the 1st layer of watermark is that length is that (expression watermark span is desirable arbitrarily from 1～2 for 32 binary numeral ³²Integer, promptly 1～4294967296), the 2nd layer of watermark is that length is 24 dyadic number value system, the length of the 3rd layer and the 4th layer watermark all is 16 binary numeral.

Step S142: time for reading is long successively from former sound signal is T _xAudio signal segment, purpose is to insert i layer watermark code in each audio signal segment.

If: the time T that writes 1 (bit) information in the original digital audio and video signals _Bit(unit: second) satisfy following condition:

0.05≤T _bit≤0.2

Wherein, for the convenience of the user, T _BitValue can also can specify by the user.Because the longest watermark code length is 32 binary digits, so for the implantation of single watermark code, the longest duration T _Code(unit: second) satisfy:

1.6≤T _code≤6.4。

Definite T _CodeValue depends on selected T _BitValue.

The watermark code that has these parameters can be implanted repeatedly (generally need at least to insert 2 times, therefore, the time of original digital audio and video signals length was at least 25 seconds) with interval at original digital audio and video signals in the duration.Therefore, the detection to watermark code is to need to guarantee on each different sections of original digital audio and video signals.Recommendation according to an international practice, the present invention are to realize in 10 seconds by minimum section duration, i.e. T _x=10 seconds.

Original voice data is pressed the time shaft direction, and when watermark was implanted, per 10 seconds repeated to implant watermark information one time, and such Redundancy Design has guaranteed the robustness of watermark, when original audio is blocked into multistage, also can ensure the extraction of watermark information.When extracting watermark, need in whole audio frequency time range, not detect, as long as continuous quadratic detects watermark information and is present in the audio file, just stop to detect, and think that original audio has watermark information.

And in a preferred embodiment, to m _bAnd T _BitThe optimal selection of occurrence will be decided according to the time and the frequency characteristic of original digital audio and video signals, promptly makes it to satisfy m as far as possible _b* T _Bit* n ≈ T _Track, n is a positive integer, T _TrackDuration for former sound signal.

Then, be T by step S143 and S144 in the time length that reads _xAudio signal segment in implant watermark code, and the audio signal segment that will implant watermark code writes to a new sound signal.

Step S143: calculate and be each (bit) position on the time shaft of former sound signal among binary watermark code wm1.It is 0 that the time shaft of former sound signal is exactly with the starting position, according to the duration of former sound signal since 0 time vector of the end of former sound signal extremely.That is to say, be to determine that among the watermark code wm1 each is inserted in the former sound signal with the particular location of time representation, promptly implants constantly.

Step S144: among the watermark code wm1 each implantation constantly, do Fourier (Fourier) conversion with implanting former sound signal constantly, time domain is transformed to frequency field, implant corresponding watermark bit by following watermarking algorithm formula (1) and (2).

The implantation of watermark code is not an amplitude of revising sound signal on time domain simply, but at transform domain the coefficient of frequency of signal is carried out analytical calculation.Former sound signal is equivalent to communication channel, watermark is equivalent to the signal by Channel Transmission, according to the spread spectrum communication principle, each of watermark code is distributed in a plurality of frequency coefficients of original audio signal, it is very little and can not arbitrarily detect to be added in signal energy on each coefficient of frequency.

Wherein, original digital audio and video signals is expressed as: x _n, n=0,1,2 ..., N-1, wherein N is the sampling number of discrete signal.The sample frequency of former sound signal is F _s, common voice signal sequence is a wav file, as mentioned above, watermark will be implanted former sound signal x as an extra _nIn.

For i frequency band (f in former sound signal _i, g _i) watermark of middle insertion i layer, at each implantation moment of watermark, i frequency band (f in the former sound signal _i, g _i) the respectively corresponding amplitude A of bound frequency values _Fi, A _GiCalculate by following formula:

$A_{\mathrm{fi}}=\sqrt{{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\cos \left(\frac{{2\mathrm{\&pi;f}}_i}{F_s}\right)x_n\right]}^2+{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\sin \left(\frac{2{\mathrm{\&pi;f}}_i}{F_s}\right)x_n\right]}^2}---\left(1\right)$

$A_{\mathrm{gi}}=\sqrt{{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\cos \left(\frac{{2\mathrm{\&pi;g}}_i}{F_s}\right)x_n\right]}^2+{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\sin \left(\frac{2{\mathrm{\&pi;g}}_i}{F_s}\right)x_n\right]}^2}---\left(2\right)$

Wherein, N represents the sampled point number of former sound signal.

X in the formula _nBe former sampled audio signal value sequence, F _sBe sample frequency, f _i, g _iBe respectively selected i frequency band (f _i, g _i) the bound frequency values.

If the binary digit of implanting is 1, and Then new value is calculated as follows: Otherwise new value is got the A that calculates _FiAnd A _Gi

When watermark was implanted, the data of in batches reading in former sound signal were to internal memory, and watermarking algorithm calculates and analyzes it, is written to then in the new sound signal in batches, and promptly the audio file with watermark is two different files with former audio file.

Similarly, if the binary digit of implanting is 0, and

Then new value is calculated as follows: Otherwise new value is got the A that calculates _FiAnd A _Gi

γ _sBe user's controlled variable, come original signal is revised by its value:

1＜γ _s＜10。

By user's controlled variable original signal is repaiied, the choosing of user's controlled variable should guarantee that watermark code can implant safely, guarantees that again watermark code can read reliably.

Allow the normal distortion of signaling channel medium frequency characteristic, in fact do not change the amplitude ratio, because the frequency of selecting is in narrow relatively frequency range.

Certainly, the present invention can also speed up the check code of implanting 16 (scale-of-two) after watermark code when implanting watermark code, be used for guaranteeing the correct extraction of watermark code.When definite i layer watermark code, the randomizer function produces any integer value (being watermark code) that algorithm predefine is good; Then, watermark code is as the seed of randomizer function, when the identical kind period of the day from 11 p.m. to 1 a.m of input, can get an identical random number sequence, this random number sequence is used for calculating the cyclic redundancy check (CRC) code of watermark code, 16 of verification code lengths, meeting and watermark code together are implanted in the sound signal when watermark is implanted.For example: suppose that we will be 201 seconds audio file implantation ground floor watermark to a first duration, watermark value wm1=8564423, it is converted into scale-of-two is: 1,000 0,010 1,010 1,110 1,100 0111, because ground floor watermark word length is 32, so fill 80 in front, the watermark code that will implant is at last: 0,000 0,000 1,000 0,010 10,101,110 1,100 0111.If total duration of audio file is 201 seconds, watermark code adds 48 altogether of check codes, if get T _Code=2.01 seconds, obtain T _Bit=0.042 second, do not satisfy T _BitCondition is so get T _Code=4.02 seconds, obtain T _Bit=0.084 second, parameters all satisfied the watermark implantation condition like this, pressed time shaft and former audio file is implanted 48 watermark (preceding 32 are watermark code, and back 16 is check code) got final product.

Watermark detection algorithms is the inverse process that algorithm is implanted in watermark, for detecting the adjacent position (bit) of watermark code, has designed two frequency detectors.The value of supposing next adjacent position is:

B _i, i=0,1...m _b-1, m _bBe the length of the binary code of watermark value

During watermark extracting, specify the watermark code of extracting in this frequency band by input frequency band parameter by the user.If the user does not specify, then successively from the acquiescence 4 frequency bands (the 1st frequency band (f ₁, g ₁), the 2nd frequency band (f ₂, g ₂), the 3rd frequency band (f ₃, g ₃) and the 4th frequency band (f ₄, g ₄)) in carry out watermark extracting action successively, duration and frequency band corresponding watermark code length by to the audio file that has watermark can calculate T _Bit, by taking out T _BitThe sound signal of the start time point in the time interval and concluding time point is transformed into frequency field, can calculate the frequency band (f at watermark code place _i, g _i).

Calculate frequency band (f _i, g _i) in the respective magnitudes of watermark code place sound signal constantly, as shown in the formula:

${\hat{A}}_{\mathrm{fi}}=\sqrt{{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\cos \left(\frac{{2\mathrm{\&pi;f}}_i}{F_s}\right)y_n\right]}^2+{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\sin \left(\frac{2{\mathrm{\&pi;f}}_i}{F_s}\right)y_n\right]}^2}---\left(3\right)$

${\hat{A}}_{\mathrm{gi}}=\sqrt{{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\cos \left(\frac{{2\mathrm{\&pi;g}}_i}{F_s}\right)y_n\right]}^2+{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\sin \left(\frac{2{\mathrm{\&pi;g}}_i}{F_s}\right)y_n\right]}^2}---\left(4\right)$

Wherein, N represents audio sample point sum.

Here y _nRepresent new sound signal (having watermark information), the amplitude ratio is:

${\mathrm{\&Delta;}}_i=\frac{{\hat{A}}_{f1}}{{\hat{A}}_{\mathrm{gi}}}$

Wherein, $B_i=\left\{\begin{array}{c}1,({\mathrm{\&Delta;}}_i\&GreaterEqual;1)\\ 0,({\mathrm{\&Delta;}}_i<1)\end{array}\right.$

To new sound signal y _nThe output result who handles is a serial bit sequence:

{B _i}，i＝0，1...，m _b-1。

The Δ that calculates _iValue will be as the reliable criterion of watermark code place value.

Each is logical zero or 1 to determine to calculate watermark code by frequency detecting, the voice data that has watermark code is traveled through, obtain meeting the watermark information of frequency band watermark code length, must carry out self checking to this information, abandon the watermark information of imperfect coupling, up to obtaining complete errorless watermark code.

Because each frequency band watermark code length fixes, and the duration of audio file also fixes, and can calculate each duration and implantation or extraction time of each frequency band watermark code.

The above only is preferred embodiment of the present invention, not in order to restriction the present invention, all any modifications of being done within the spirit and principles in the present invention, is equal to and replaces and improvement etc., all should be included within protection scope of the present invention.

Claims (10)

1. a DAB multilayer watermark method for implantation is characterized in that, comprises step:

The i frequency band of a, definite insertion i layer watermark in former sound signal;

B, when inserting the watermark of i layer, judge whether i-1 is 0, if be not 0, change step c over to, otherwise continuation judges whether implanted the watermark of i-1 layer in the i-1 frequency band, if then change step c over to;

C, insertion i layer watermark in the i frequency band of former sound signal;

Wherein, 4 〉=i 〉=1, and i is a positive integer.

2. according to the described DAB multilayer of claim 1 watermark method for implantation, it is characterized in that step c comprises:

C1, determine the watermark code of i layer watermark;

C2, read the audio signal segment of schedule time length successively, in each audio signal segment, implant watermark code respectively, the audio signal segment of implanting watermark code is written to a new sound signal.

3. according to the described DAB multilayer of claim 2 watermark method for implantation, it is characterized in that step c2 comprises:

C21, implantation moment of calculating each correspondence on the time shaft of former sound signal in binary watermark code;

C22, the former sound signal that will implant are constantly done Fourier transform, and time domain is transformed to frequency field, calculate the new amplitude of i frequency band after the watermark code of implanting the watermark of i layer in frequency field.

4. according to the described DAB multilayer of claim 3 watermark method for implantation, it is characterized in that, comprise in the calculation procedure of the new amplitude of i frequency band described in the step c22:

Each implants constantly the upper limiting frequency f of i frequency band in the former sound signal watermark code of c221, the watermark of calculating i layer _iWith lower frequency limit g _iCorresponding respectively amplitude A _Fi, A _Gi:

$A_{\mathrm{fi}}=\sqrt{{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\cos \left(\frac{{2\mathrm{\&pi;f}}_i}{F_s}\right)x_n\right]}^2+{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\sin \left(\frac{2{\mathrm{\&pi;f}}_i}{F_s}\right)x_n\right]}^2},$

$A_{\mathrm{gi}}=\sqrt{{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\cos \left(\frac{{2\mathrm{\&pi;g}}_i}{F_s}\right)x_n\right]}^2+{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\sin \left(\frac{2{\mathrm{\&pi;g}}_i}{F_s}\right)x_n\right]}^2},$

If the watermark code binary digit that c222 implants is 1, and

The upper limiting frequency f of i frequency band in the time of then should implanting constantly _iWith lower frequency limit g _iThe corresponding respectively following calculating of new amplitude:

Otherwise the upper limiting frequency f of i frequency band _iWith lower frequency limit g _iDui Ying new amplitude equals the amplitude A that calculates among the step c221 respectively _FiAnd A _Gi

If the watermark code binary digit that c223 implants is 0, and

The upper limiting frequency f of i frequency band in the time of then should implanting constantly _iWith lower frequency limit g _iThe corresponding respectively following calculating of new amplitude:

Otherwise the upper limiting frequency f of i frequency band _iWith lower frequency limit g _iDui Ying new amplitude equals the amplitude A that calculates among the step c221 respectively _FiAnd A _Gi

Wherein, N represents the sampled point number of former sound signal; x _nBe former sampled audio signal value sequence, F _sBe sample frequency; γ _sBe user's controlled variable.

5. according to the described DAB multilayer of claim 4 watermark method for implantation, it is characterized in that 1＜γ _s＜10.

6. according to the described DAB multilayer of claim 1 watermark method for implantation, it is characterized in that, describedly judge whether implanted the watermark of i-1 layer in the i-1 frequency band, be by the i-1 frequency band being carried out the watermark extracting operation, if the discovery watermark is promptly thought and has been implanted the watermark of i-1 layer in the i-1 frequency band.

7. according to the described DAB multilayer of claim 1 watermark method for implantation, it is characterized in that, watermark code be one between 16 to 32 binary numeral.

8. according to any one described DAB multilayer watermark method for implantation of claim 1-7, it is characterized in that the 1st frequency band is 500Hz-600Hz; The 2nd frequency band is 550Hz-650Hz; The 3rd frequency band is 700Hz-800Hz; The 4th frequency band is 750Hz-850Hz.

9. a DAB multilayer watermark extracting method is characterized in that, comprises step:

To the frequency band of user's appointment, or 4 frequency bands from giving tacit consent to, carry out the watermark extracting operation;

Determine the time information of watermark code in sound signal in the i frequency band by two frequency detectors;

Convert sound signal to frequency field by Fourier transform, in frequency field, calculate the watermark code place amplitude of i frequency band constantly;

By i frequency band upper limiting frequency f _iCorresponding amplitude With lower frequency limit g _iCorresponding amplitude

Amplitude ratio:

Obtain the binary sequence B of watermark code _i, wherein $B_i=\left\{\begin{array}{c}1,({\mathrm{\&Delta;}}_i\&GreaterEqual;1)\\ 0,({\mathrm{\&Delta;}}_i<1)\end{array}\right..$

10. a DAB multilayer watermark extracting method is characterized in that, the amplitude of i frequency band upper limiting frequency correspondence The amplitude corresponding with lower frequency limit

Be respectively:

${\hat{A}}_{\mathrm{fi}}=\sqrt{{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\cos \left(\frac{{2\mathrm{\&pi;f}}_i}{F_s}\right)y_n\right]}^2+{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\sin \left(\frac{2{\mathrm{\&pi;f}}_i}{F_s}\right)y_n\right]}^2};$

${\hat{A}}_{\mathrm{gi}}=\sqrt{{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\cos \left(\frac{{2\mathrm{\&pi;g}}_i}{F_s}\right)y_n\right]}^2+{\left[\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\sin \left(\frac{2{\mathrm{\&pi;g}}_i}{F_s}\right)y_n\right]}^2};$

Wherein, y _nThe sound signal of expression band watermark, N represents the sampled point number of former sound signal, F _sBe sample frequency.

Images