CN111755018B - Audio hiding method and device based on wavelet transform and quantized embedded key - Google Patents

Abstract

The invention discloses an audio concealment method and device based on wavelet transform and quantization embedding key. The method includes the following steps: acquiring carrier audio and secret audio, performing secondary wavelet transform on the carrier audio, and obtaining the position of concealed information to be embedded; Extract the extremum and extremum coordinates of the secret audio, and obtain the concealed information according to the extremum and extremum coordinates; according to the location of the concealed information to be embedded, embed the concealed information into the carrier audio to obtain the first audio; use the quantization method to embed the key The carrier audio that does not contain the concealed information is used to obtain the second audio, and the first audio and the second audio are concatenated to obtain the transmission audio. The present invention performs secondary wavelet transformation on the carrier audio to embed the extremum of the secret audio and its differential sequence, so as to achieve a balance in robustness and transparency; in addition, the key watermark is embedded into the carrier audio by using a quantization method , to further enhance the security of audio transmission, and can be widely used in the field of information security.

Description

Audio hiding method and device based on wavelet transform and quantized embedded key

technical field

The invention relates to the field of information security, in particular to an audio hiding method and device based on wavelet transform and quantized embedded key.

Background technique

With the rapid development of multimedia technology, it has brought a lot of entertainment and convenience to people, but it also exposed serious security problems. Illegal interception of information in network transmission leads to information leakage is also an important information security issue, especially confidential information in military, financial, commercial and other fields, "protecting information" and "covering communication" are also one of the research hotspots in the field of network information security .

The audio hiding system is to hide a piece of secret audio in another carrier without destroying the information of the original carrier. After the carrier carrying the secret is transmitted, the secret audio embedded in it can still be extracted. Concealment technology Compared with "self-scrambling" encryption technology, concealment technology has more advantages in protecting secret audio. Compared with files such as text and pictures, audio files contain irreplaceable and unique personal characteristics such as pitch, timbre, and speech rate, so research on hiding audio is more meaningful. However, the data volume of secret audio information is large, the storage method of data is more complex and there are some redundant information, which makes the embedding capacity large and makes hiding difficult. But on the other hand, the complexity of audio data also shows that audio itself can be used as a carrier to hide audio.

In the development history of audio concealment technology, there are four classic audio concealment techniques: least significant bit method, spread spectrum method, phase coding method and echo coding method. These four methods are simple to operate and have good transparency. Although the hidden system is not easily affected by intentional attacks during transmission, it is vulnerable to unintentional attacks, such as: TSM attack, noise attack, random clipping attack, point attack, compression attack, noise reduction attack, etc. These four classic methods are relatively poor in resisting various attacks, and are easy to be changed by these unintentional attacks, making it difficult to extract the secret information inside. Wavelet analysis was originally used for compression and denoising, but with the development of computer technology and the improvement of mathematical theory, many achievements have been made in signal processing. Wavelet analysis has local characteristics in both time domain and frequency domain, which is different from Fourier transform which affects the entire time domain, so it is more suitable for dynamic concealment technology.

Contents of the invention

In order to solve one of the above technical problems, the object of the present invention is to provide an audio hiding method and device based on wavelet transform and quantized embedded key.

The technical scheme adopted in the present invention is:

An audio hiding method based on wavelet transform and quantized embedded key, comprising the following steps:

Obtain carrier audio and secret audio, perform secondary wavelet transform on the carrier audio, and obtain the position to be embedded with hidden information;

Extract the extremum and extremum coordinates of the secret audio, and obtain hidden information according to the extremum and extremum coordinates;

Embedding the concealed information into the carrier audio according to the position of the concealed information to be embedded to obtain the first audio;

A quantization method is used to embed the key into the carrier audio that does not contain concealed information to obtain the second audio, and to concatenate the first audio and the second audio to obtain the transmission audio.

Further, the extracting the extremum and extremum coordinates of the secret audio, and obtaining concealed information according to the extremum and extremum coordinates include:

extract the extrema and extrema coordinates of the secret audio;

Extract the difference sequence of the extreme value coordinates, and perform stretching transformation on the difference sequence of the extreme value and the extreme value coordinates;

The difference sequence of extremum and extremum coordinates after telescopic transformation is reconstructed by Haar wavelet to obtain hidden information.

Further, the second wavelet transform is performed on the carrier audio to obtain the position of the hidden information to be embedded, including:

Carrier audio is decomposed by two-level Haar wavelets to obtain the first high-frequency coefficient, the first intermediate-frequency coefficient and the first low-frequency coefficient;

The Haar wavelet first-level decomposition is performed on the first intermediate-frequency coefficient to obtain the second high-frequency coefficient and the second low-frequency coefficient, and the position of the second low-frequency coefficient is used as the position to be embedded with hidden information.

Further, the embedding the concealed information into the carrier audio according to the position of the concealed information to be embedded to obtain the first audio includes:

Embedding the hidden information into the second low-frequency coefficient to obtain the third low-frequency coefficient;

The second high-frequency coefficient and the third low-frequency coefficient are subjected to the reconstruction of the second Haar wavelet transform to obtain the third intermediate-frequency coefficient;

The first audio frequency is obtained by performing Haar wavelet first-level reconstruction on the first high-frequency coefficient, the third intermediate-frequency coefficient and the first low-frequency coefficient.

Further, the method of embedding the key watermark into the carrier audio that does not contain concealed information using a quantization method to obtain the second audio, and concatenating the first audio and the second audio to obtain the transmission audio includes:

Obtain the length of the extremum sequence as the key, and convert the key into a binary code;

performing discrete cosine transform on the carrier audio that does not contain concealed information to obtain a fourth low-frequency coefficient and a fourth high-frequency coefficient;

Embedding the binary coded quantization into the fourth low-frequency coefficient to obtain the fifth low-frequency coefficient;

Inverse discrete cosine transform is performed on the fifth low-frequency coefficient and the fourth high-frequency coefficient to obtain the carrier audio containing the key as the second audio;

Connect the first audio and the second audio in series to obtain transmission audio.

Further, the following steps are also included:

Obtain the transmitted audio, and extract the key from the transmitted audio by discrete cosine transform;

Combining secret key and quadratic wavelet transform to obtain secret audio from transmitted audio.

Further, said adopting the discrete cosine transform to extract the key from the transmission audio includes:

Discrete cosine transform is performed on the transmission audio containing the key to obtain the sixth low-frequency coefficient;

A binary code is obtained from the sixth low-frequency coefficient, and the binary code is converted into a key.

Further, the combination of key and secondary wavelet transform to obtain secret audio from transmission audio includes:

Obtaining a third audio containing concealed information from the transmission audio according to the key, and performing Haar wavelet secondary decomposition on the third audio to obtain a fourth intermediate frequency coefficient;

Perform Haar wavelet first-level decomposition on the fourth intermediate frequency coefficient to obtain the seventh low frequency coefficient;

obtaining concealed information from the seventh low-frequency coefficient, the concealed information including extremums and differential sequences of extremum coordinates;

Perform first-order spline interpolation on the extreme value and the difference sequence of the extreme value coordinates to obtain the secret audio.

Another technical scheme adopted in the present invention is:

An audio concealment device based on wavelet transform and quantized embedding key, an audio transform module, used to obtain carrier audio and secret audio, perform secondary wavelet transform on the carrier audio, and obtain the position of hidden information to be embedded;

The extremum extraction module is used to extract the extremum and extremum coordinates of the secret audio, and obtain hidden information according to the extremum and extremum coordinates;

The audio embedding module is used to embed the concealed information into the carrier audio according to the position of the concealed information to be embedded to obtain the first audio;

The key embedding module is used to embed the key into the carrier audio that does not contain hidden information by using a quantization method to obtain the second audio, and concatenate the first audio and the second audio to obtain the transmission audio.

Another technical scheme adopted in the present invention is:

An audio concealment device based on wavelet transform and quantized embedded key, comprising:

at least one processor;

at least one memory for storing at least one program;

When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

The beneficial effect of the present invention is: the present invention does secondary wavelet transform to carrier audio frequency, thereby finds the position that imperceptibility and robustness are balanced, to embed extremum of secret audio frequency and difference sequence thereof, make in robustness and robustness In addition, in order not to open a secret channel to transmit the key, the quantization method is used to embed the key watermark into the carrier audio to further enhance the security of audio transmission.

Description of drawings

Fig. 1 is a schematic structural diagram of an audio hiding system in an embodiment;

Fig. 2 is the flowchart of secret audio embedding process and key watermark embedding process in the embodiment;

Figure 3 is a schematic diagram of the comparison between the carrier audio containing concealed information and the original carrier audio in the embodiment;

Fig. 4 is the flowchart of key watermark extraction process and secret audio extraction process in the embodiment;

Fig. 5 is the comparison chart of the secret audio obtained by first-order spline interpolation and the original secret audio in the embodiment;

Fig. 6 is a structural block diagram of an audio hiding device based on wavelet transform and quantization embedding key in an embodiment.

Detailed ways

Embodiments of the present invention are described in detail below, examples of which are shown in the drawings, wherein the same or similar reference numerals designate the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the figures are exemplary only for explaining the present invention and should not be construed as limiting the present invention.

In the description of the present invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc. indicated orientations or positional relationships are based on the orientations or positional relationships shown in the drawings, and are only In order to facilitate the description of the present invention and simplify the description, it does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present invention.

In the description of the present invention, several means one or more, and multiple means more than two. Greater than, less than, exceeding, etc. are understood as not including the original number, and above, below, within, etc. are understood as including the original number. If the description of the first and second is only for the purpose of distinguishing the technical features, it cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features relation.

In the description of the present invention, unless otherwise clearly defined, words such as setting, installation, and connection should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above words in the present invention in combination with the specific content of the technical solution.

This embodiment is based on wavelet transform and quantized audio hiding system with embedded key. The structural diagram is shown in Figure 1, which is divided into a secret information sending end and a secret information receiving end, wherein the secret audio sending end includes a secret information embedding process and a key In the watermark embedding process, the secret audio receiver includes the key watermark extraction process and the secret information extraction process.

The specific methods involved in the four processes are as follows:

The secret information embedding process involves a secondary wavelet transform-based information embedding method, using another segment of audio as a hidden carrier for secret audio, including the following steps 1.1-step 1.6:

Step 1.1: Extract the extrema and extremum coordinates of the secret audio.

Let the length of the secret audio be n×1, and record the audio sequence as secret={secret _i |0≤i≤n}. Then the extreme value coordinate sequence Loc of the audio signal is:

Loc＝{i+1|temp _i -temp _i-1 ＝-2or temp _i -temp _i-1 ＝2,i＝1,2,…,n-1}

Where: temp=sign(diff(secret)).

Then the extreme value sequence optima is:

optima={secret _i |i∈Loc}

Step 1.2: Extract the difference sequence of the extremal coordinates.

The differential sequence Loc _diff of the extreme coordinates is as follows:

Loc _diff ＝{Loc ₁ }∪{Loc _i -Loc _i - ₁ |i＞1}

Step 1.3: Perform Haar wavelet first-level reconstruction on the extremum and its coordinate difference sequence after stretching transformation to obtain hidden information.

The extremum sequence optima is changed to the original x times, and the coordinate difference sequence Loc _diff is changed to the original y times, which are used as the low-frequency coefficients and high-frequency coefficients of the wavelet reconstruction respectively, and a hidden sequence S is obtained by pairwise matching reconstruction.

Among them, the parameters x, y are usually adaptively adjusted according to the intensity of the audio, and the transformation factors x, y are used as two keys.

Step 1.4: Perform secondary Haar wavelet decomposition on the carrier audio.

Step 1.4.1: Perform second-level Haar wavelet decomposition on the carrier audio.

The volume audio signal is recorded as carrier, and Haar wavelet is used to decompose the carrier signal by two-level wavelet first, and the decomposed low frequency coefficient cA10, high frequency coefficient cD10 and intermediate frequency coefficient cD10 are obtained.

Step 1.4.2: Perform Haar wavelet first-level decomposition on the intermediate frequency coefficients generated in step 1.4.1.

Then, Haar wavelet is used to decompose the intermediate frequency coefficient cD10 to obtain the low frequency similarity coefficient cA11 and high frequency coefficient cD11.

Step 1.5: Embed concealed information in the low-frequency coefficient cA11 generated in step 1.4.2.

Calculate the length L _cA11 of the low-frequency coefficient cA11 of the quadratic wavelet transform, and at the same time calculate the length L _optima of the secret audio extremum sequence optima.

If L _cA11 <2×L _optima , it indicates that the embedding position is not enough to store the hidden sequence, and the carrier audio needs to be lengthened, that is, jump back to step 1.4.

If L _cA11 ≥ 2×L _optima , replace the front part of cA11 with the covert sequence S to obtain a new low-frequency coefficient cA11 ^new .

Wherein, the length L _optima of the extremum sequence optima is used as a key.

Step 1.6: Reconstruct the wavelet coefficients containing hidden information by secondary Haar wavelet transform.

Step 1.6.1: Perform Haar wavelet first-level reconstruction on high-frequency coefficients and low-frequency coefficients containing hidden information.

Reconstruct the unchanged high-frequency coefficient cD11 and cA11 ^new embedded with hidden information into a new intermediate-frequency coefficient cD1o ^new ;

Step 1.6.2: Perform Haar wavelet secondary reconstruction on the low-frequency, high-frequency and intermediate-frequency coefficients containing hidden information.

Then carry out Haar wavelet secondary reconstruction on cD10 ^new and unchanged cD20 and cA10 to get the audio carrier ^new embedded with secret.

The key watermark embedding process involves a hidden algorithm for quantifying the embedding key, and the specific quantifying embedding process includes steps 2.1-step 2.6:

Step 2.1: Convert the key key={L _optima , x, y} into a binary code.

Use L1 bit values to store the binary code of L _optima , that is, the decimal L _optima is converted into binary 0, 1 code;

Use L2 bits to store the binary code of a, where the first bit value is taken as sign(a-1), if a<1, then the last L2-1 bit value stores the code of 1/a, otherwise it is stored directly the code of a;

Similarly, use L3 bits to store the binary code of b, where the first bit value is taken as sign(b-1), if b<1, then the last L3-1 bit value stores the code of 1/b, otherwise Directly store the encoding of b;

Concatenate the three strings of binary sequences to obtain the binary sequence W. Among them, L1, L2, and L3 are determined in advance based on experience.

Step 2.2: Judging whether the carrier that does not contain hidden information can store binary codes.

Since the carrier containing concealed signal is carrier ^new [1:16×L _optima ], the carrier not containing concealed information is carrier ^new [16×L _optima +1:end], which is abbreviated as carrier1.

Among them, end means to get the last component of the vector, the same below.

If L _carrier1 <(L1+L2+L3)*num, it means that the embedding position is not enough to store the binary sequence W, and it means that the carrier audio needs to be lengthened before the code can be stored.

If L _carrier1 ≥ (L1+L2+L3)*num, it means that the embedding position can store the binary sequence W, which means that the carrier audio is enough to store the binary code.

Among them, L _carrier1 represents the length of carrier1; num is the length of each audio segment, which is determined in advance based on experience.

Step 2.3: Discrete cosine transform is performed on the carrier audio without concealed information.

Intercept carrier ^new [end-(L1+L2+L3)*num+1: end], abbreviated as carrier2.

Carrier2 is divided into frames, divided into L1+L2+L3 frames, and each frame has num signal points.

Discrete cosine (DCT) transform is performed on each frame to obtain num expansion points, and half of them with lower frequencies are recorded as low-frequency coefficients; half of them with higher frequencies are recorded as high-frequency coefficients.

The DCT transformation is performed on each frame of audio in a loop until the DCT decomposition of the L1+L2+L3 frame audio is completed.

Step 2.4: Quantize and embed binary coding on the low-frequency coefficients transformed by DCT in each frame.

Embed the i-th watermark w _i into the i-th frame of audio, record the i-th frame of audio as the low-frequency coefficient sequence F _i of DCT transformation, and perform watermark embedding for any low-frequency coefficient fi _i ∈ F _i , the rules are as follows:

If w _i =0, the low-frequency coefficients are transformed into Q(f _ij )=k*round(fi _j /k);

If w _i =1, the low-frequency coefficients are transformed into Q(f _ij )=k*round(fi _j /k)+k/2;

Wherein, the quantization factor k is a positive even number determined in advance according to empirical values.

Circulate to embed the watermark in the low-frequency coefficients fi _j of F _i until all the low-frequency coefficients fi _j are embedded with the watermark w _i .

Loop embedding watermarks in the low-frequency coefficients of each frame of audio until all watermarks are embedded.

Step 2.5: Perform DCT inverse transform on the quantized DCT transformed coefficients to obtain the carrier signal containing the key watermark.

Step 2.6: Concatenate the carrier audio carrying the concealed information and the carrier audio carrying the key watermark to obtain the transmission audio carrier ^tsm .

The key watermark extraction process involves a method of quantifying and extracting keys, which is the inverse process of the key embedding process, specifically quantified extraction steps 3.1-step 3.3:

Step 3.1: Perform DCT transformation on the transmission audio containing the key.

Extract the transmission audio containing the key: counting from the tail, the length is (L1+L2+L3)*num, that is, carrier ^tsm [end-(L1+L2+L3)*num+1, end], recorded as carrier3;

Divide the audio carrier3 into L1+L2+L3 frames with a length of num, and perform DCT transformation on each frame of audio signal;

Step 3.2: Threshold quantization to extract binary codes from DCT transformed low-frequency coefficients.

Extract the i-th binary code w _i from the low-frequency coefficient set F _i of the i-th frame, namely

If #(abs(f'-round(f'/num)×num)≤k/4)≥num/2, then w _i =0;

If #(abs(f'-round(f'/num)×num)≥k/4)>num/2, then w _i =1;

Among them, #(·), abs(·), round(·) are counting function, absolute value function and rounding function respectively.

Step 3.3: Convert the binary code into a key.

For the previous L1 codes, directly convert these L1 binary numbers into decimal key m;

For the middle L2 codes, if the first code is 0, it means that x≤1, and the last L2-1 binary codes are converted into decimal z, so that x=1/z is obtained. Otherwise, it means x>1, and then convert the last L2-1 binary codes into decimal key x;

Similarly, for the last L3 codes, if the first code is 0, it means y≤1, and the last L3-1 binary codes are converted into decimal z, so that y=1/z is obtained. Otherwise, it means y>1, and then convert the last L3-1 binary codes into decimal key y;

Finally, the key key={m, x, y} is obtained.

The secret audio extraction process is the reverse process of the secret audio embedding process, that is, the hidden information in the transmission audio needs to be extracted according to the key, and the specific steps of the extraction method are as follows:

Step 4.1: Perform secondary Haar wavelet decomposition on the transmitted audio.

Step 4.1.1: Perform Haar wavelet secondary decomposition on the transmitted audio.

According to the length L _optima of the key extremum sequence, it is determined that the transmission audio part containing concealed information is carrier ^tsm [1: L _optima × 16], which is abbreviated as carrier4.

Use Haar wavelet to do two-stage wavelet transform on the audio carrier4 containing hidden information, and obtain the decomposed low-frequency coefficient cA10, high-frequency coefficient cD20 and intermediate-frequency coefficient cD10.

Step 4.1.2: Perform Haar wavelet first-level decomposition on the intermediate frequency coefficients generated in step 4.1.1.

Then perform a first-level wavelet decomposition on the intermediate frequency coefficient cD10 ^new , and extract the decomposed low frequency coefficient cA11 ^new and high frequency coefficient cD11.

Step 4.2: Extract hidden information from the low-frequency coefficients cA11 ^new generated in Step 4.1.2.

The low-frequency coefficient cA11 ^new is subjected to a first-level discrete Haar wavelet transform to obtain low-frequency coefficients and high-frequency coefficients, which correspond to the extremum sequence and extremum coordinate difference sequence of the stretching transformation respectively.

Then use the key expansion factor x, y to obtain the extreme value sequence optima of the secret audio and the differential sequence Loc _diff of its coordinates, and then obtain the coordinate sequence Loc of the extreme value by the following formula:

When i=1, Loc _i =Loc_diff ₁ ; when i>1, Loc _i =Loc_diff _i-1 +Loc _i-1 .

Step 4.3: Perform first-order spline interpolation on the extremum and its coordinate difference sequence to obtain the simulated secret audio.

The extremum sequence optima and its coordinate sequence Loc are correspondingly matched into points, and the first-order spline method is used to interpolate to obtain the simulated secret ^new of the original secret signal.

Based on the above system, this embodiment also provides an audio hiding method based on wavelet transform and quantized embedded key, including but not limited to the following steps:

Secret audio embedding process: embed the secret audio into the carrier audio, and obtain the key-carrying secret audio;

Key watermark embedding process: quantify the key and embed it into the carrier audio to obtain audio that can be transmitted from the secret information sender to the secret information receiver;

Key watermark extraction process: extract the key and the audio carrying the secret from the transmitted audio to assist the secret audio extraction process to extract the secret audio;

Secret audio extraction process: Use the key to obtain simulated secret audio from the audio carrying the secret.

In this embodiment, the carrier audio uses the audio content in Chinese as "The road is long and long, I will search up and down", and the secret audio uses the English content as "It is my report from the north, which cheekily induced people to buy. " audio.

Referring to Figure 2, in the secret audio embedding process, the secret audio extremum and its position information are used to replace the original secret audio embedding to compress the embedding capacity; the carrier audio is subjected to secondary wavelet transform to obtain the position of transparency and robustness balance to obtain Embed covert information; detail steps 1.1-step 1.6:

Step 1.1: Extract the extrema and extremum coordinates of the secret audio.

Since this embodiment uses the audio signal as a secret signal, and the audio signal has a large amount of data, and there are some redundant information, it will increase the difficulty due to the increase in the embedding capacity, so the extremum sequence of the audio signal and its coordinate sequence are extracted as audio characteristics. Experiments have shown that interpretable audio signals can be restored by first-order spline interpolation using only extreme values and their coordinates.

Step 1.2: Extract the difference sequence of the extremal coordinates.

As the length of the audio signal increases, the extreme value coordinates also increase. In order to avoid negative effects on the embedding effect, a differential operation is performed on the extreme value coordinate sequence to obtain a relatively stable one that is not affected by the length of the audio signal. Sequence of extreme value coordinate differences.

Step 1.3: Perform Haar wavelet first-level reconstruction on the extremum and its coordinate difference sequence after stretching transformation to obtain hidden information.

The extremum sequence and the extremum coordinate difference sequence are respectively stretched and transformed, so that the values of the two are similar. The extremum sequence and the coordinate difference sequence are matched in pairs. In order to make the dimensions of the secret information the same, the extremum sequence and the extremum coordinate difference sequence of the stretching transformation are used as the low-frequency coefficients and high-frequency coefficients of the Haar wavelet reconstruction, and the reconstructed Obtain a covert sequence whose components have the same dimension.

Wherein, for the Haar wavelet first-level reconstruction, set the low-frequency coefficient sequence as A=[a ₁ ,a ₂ ,…,a _n ], and the high-frequency coefficient sequence as B=[b ₁ ,b ₂ ,…,b _n ], then the Haar wavelet reconstruction result is: the 2i-1th signal value is f _2i-1 =a _i +b _i ; the 2ith signal value is f _2i =a _i -b _i . In addition, the second-level Haar wavelet reconstruction described in step 1.6.1 below is to perform a first-level Haar wavelet reconstruction on the low-frequency coefficients on the basis of the first-level reconstruction.

Step 1.4: Perform secondary Haar wavelet decomposition on the carrier audio.

Step 1.4.1: Perform second-level Haar wavelet decomposition on the carrier audio.

Perform Haar wavelet secondary decomposition on the carrier audio to obtain low-frequency coefficients, intermediate-frequency coefficients, and high-frequency coefficients. Since the low-frequency coefficients represent the audio envelope, operations on them can easily have a large impact on the audio; while the high-frequency coefficients represent audio details and are easily affected by denoising attacks; in summary, the hidden signal will be embedded on the intermediate frequency coefficients .

Step 1.4.2: Perform Haar wavelet first-level decomposition on the intermediate frequency coefficients generated in step 1.4.1.

Then, the first-level wavelet decomposition is performed on the intermediate frequency coefficients to obtain low frequency coefficients and high frequency coefficients. According to wavelet theory, the low-frequency coefficients obtained in this step are more stable than the original sequence, that is, more stable than the intermediate-frequency coefficients generated in step 1.4.1. And because this coefficient does not represent the audio envelope, it can be used to store secret audio.

Wherein, the first-level decomposition of the Haar wavelet, assuming that the signal sequence is f=[f ₁ , f ₂ ,..., f _n-1 , f _n ], then the result of the Haar wavelet decomposition is: the i-th low-frequency coefficient is a _i =(f _2i-1 +f _2i )/2; the i-th high frequency coefficient is b _i =(f _2i -f _2i-1 )/2. In addition, the Haar wavelet two-level decomposition, that is, on the basis of the first-level decomposition, perform a Haar wavelet first-level decomposition on the low-frequency coefficients.

Step 1.5: Embedding hidden information in the low-frequency coefficients produced in Step 1.4.2.

Under the condition that the storage location is sufficient, replace the part before the low-frequency coefficients in step 1.4.2 with the covert sequence generated in step 1.3.

Step 1.6: Reconstruct the wavelet coefficients containing hidden information by secondary Haar wavelet transform.

Step 1.6.1: Perform Haar wavelet first-level reconstruction on high-frequency coefficients and low-frequency coefficients containing hidden information.

As the inverse step of step 1.4.2, Haar wavelet first-level reconstruction is performed to obtain intermediate frequency coefficients containing hidden information.

Step 1.6.2: Perform Haar wavelet secondary reconstruction on the low-frequency, high-frequency and intermediate-frequency coefficients containing hidden information.

As the inverse step of step 1.4.1, Haar wavelet secondary reconstruction is performed to obtain the carrier audio containing hidden information. Please refer to Figure 3, which is a comparison chart between the carrier audio embedded with hidden information and the original carrier audio. It is easy to find that the two values are extremely similar in waveform. In addition, through intuitive listening to the carrier audio embedded with hidden information, it can be found that there is only a little noise, but there is no difference in timbre, sound quality, pitch, content, etc.

Referring to Figure 2, in the key watermark embedding process, the key is quantized and embedded in the carrier audio, so that no additional secret channel is opened to transmit the key, specifically including steps 2.1-step 2.6:

Step 2.1: Convert the key into a binary code.

In the system of this embodiment, the key is only three pieces of data, and the amount of data is small, so it can be converted into binary data and embedded into the carrier audio by using quantized blind watermarking technology. The fixed-length bits are used to access the binary codes respectively, so as to quantize and extract the key. Since the length of the secret audio extremum sequence is an integer, it is directly converted from decimal to binary code; when performing scaling transformation, one of the magnification or the reciprocal of the magnification can be controlled to be an integer to obtain the stretching function, which is stored in one bit The magnitude of the magnification, and then use the method of converting decimal to binary to access the integer, that is, the magnification or the reciprocal of the magnification.

Step 2.2: Judging whether the carrier that does not contain hidden information can store binary codes.

In order not to let the embedding of the key interfere with the hidden information, it is considered to quantize the binary code of the embedded key in the carrier that does not contain the hidden information.

Step 2.3: Discrete cosine transform is performed on the carrier audio without concealed information.

In order to confirm the carrier audio containing the key after transmission, a sufficiently long audio signal is intercepted from the end of the carrier audio.

Wherein, if the signal is F={f ₀ , f ₁ , f ₂ ,..., f _N }, then the discrete cosine transform formula is:

Among them, when u=0, When u=1, />

Correspondingly, the DCT inverse transform formula is:

In the following step 4, the low-frequency coefficients of half of the DCT transformation are quantized and embedded coded. In order to improve the robustness of the quantized embedded binary code, the length of each frame of audio should be as large as possible. Experiments show that each frame signal is generally at least 32 signals.

Step 2.4: Quantize and embed binary coding on the low-frequency coefficients transformed by DCT in each frame.

Similar to the classical watermarking method, if the code is 0, the low-frequency coefficients in each frame are converted into adjacent even numbers; if the code is 1, the low-frequency coefficients in each frame are converted into adjacent odd numbers. The quantization factor is used to quantize and embed the binary code. The larger the quantization factor, the stronger the perceptibility and robustness of the watermark, and vice versa. Therefore, a balance of perceptuality and robustness is obtained by taking an appropriate quantization factor.

Compared with high-frequency coefficients, quantization embedding binary codes in low-frequency coefficients is more robust.

Step 2.5: Perform DCT inverse transform on the quantized DCT transformed coefficients to obtain the carrier signal containing the key watermark.

Step 2.6: Concatenate the carrier audio carrying the concealed information with the carrier audio carrying the key watermark to obtain the transmission audio.

In the key watermark extraction process, threshold quantization is used to extract the key watermark to enhance the robustness of the algorithm; after transmission through the channel, the key needs to be extracted first, as shown in Figure 4, which is the reverse process of the key embedding process, and the specific threshold quantization Extraction Process Step 3.1-

Step 3.3:

Step 3.1: Perform DCT transformation on the transmission audio containing the key.

Since the key is embedded from the end in the process of key watermark embedding, it is also counted from the end here, and the extracted length is (L1+L2+L3)*num, and then divided into frames, each frame is num signal values , decompose the num signal values by DCT transformation to obtain low-frequency coefficients and high-frequency coefficients, wherein the low-frequency coefficients are embedded with binary codes.

Step 3.2: Threshold quantization to extract binary codes from DCT transformed low-frequency coefficients.

For any frame, according to the rules of binary coding to embed low-frequency coefficients, one or even all of the low-frequency coefficients should be found. If the low-frequency coefficient is an integer multiple of the quantization factor, the embedded code is 0; if the low-frequency coefficient is the quantization factor An integer multiple of , plus 0.5 times, means that the embedded code is 1. However, considering that there may be unintentional attacks during the transmission process, the low-frequency coefficients have certain fluctuations, so threshold quantization is used, that is, the low-frequency coefficients are classified into the category that is closer to the integer multiple of the quantization factor and the integer multiple of the quantization factor is added. For the class that is 0.5 times closer, the number of low-frequency coefficients in each class is calculated separately, and the code corresponding to the class with a larger number is the final code. Obviously, using threshold quantization can further improve the robustness of the algorithm.

Step 3.3: Convert the binary code into a key.

When the key is converted to binary encoding, the bit length of each key has been fixed, so it is also easy to obtain the encoding corresponding to each string of bits here. Then, according to the rules of converting the key to the binary code, the inverse rule is determined, and the rule of converting the binary code into the key can be obtained.

In the secret audio extraction process, first-order spline interpolation is used to interpolate the extreme value and position of the secret audio; after having the key, the secret audio simulated in the transmitted audio is extracted according to the key, as shown in Figure 4, the specific steps of the extraction algorithm 4.1 - Step 4.3:

Step 4.1: Perform secondary Haar wavelet decomposition on the transmitted audio.

Step 4.1.1: Perform Haar wavelet secondary decomposition on the transmitted audio.

Step 4.1.2: Perform Haar wavelet first-level decomposition on the intermediate frequency coefficients generated in step 4.1.1.

Step 4.2: Extract hidden information from the low-frequency coefficients produced in Step 4.1.2.

When embedding concealed information, in order to make each information dimension the same, Haar wavelet first-level reconstruction is performed. Here, Haar wavelet first-level decomposition is required to obtain the extremum sequence of scaling transformation and the difference sequence of extremum coordinates. .

Then the difference sequence of the extreme value coordinates is transformed into the extreme value coordinates.

Because it may be affected by noise, the extreme value coordinates are not integers, but it does not affect the final analog audio, so no further processing is required.

Step 4.3: Perform first-order spline interpolation on the extremum and its coordinate difference sequence.

Although the third-order spline interpolation is more rounded than the first-order spline interpolation, and the simulated sound is more real and interpretable, but it cannot achieve good results in the experiment, because the third-order spline in the data When the spacing is large, the result of interpolation will differ greatly from the original audio, which will affect the overall shape of the audio. Therefore, in the experiment, first-order interpolation is used to perform first-order spline interpolation on the extremum and its position sequence. Refer to Figure 5, which is a comparison chart between the simulated audio obtained by using first-order spline interpolation and the original secret audio, and it is easy to find the secret of the simulation. The profile of the audio is strikingly similar to that of the original secret audio. In addition, if you directly compare the analog audio and the original secret audio from the listening point of view, you will find that there is no difference in timbre, pitch, and content between the two.

To sum up, the method of this embodiment, compared with the existing technology, has at least the following beneficial effects:

After pairing the extremum of the secret audio and its coordinate information, the simulated secret audio can be obtained by first-order spline interpolation. Therefore, only the extremum of the secret audio and its coordinate information need to be embedded in the secret audio embedding, so as to achieve the effect of compressing the audio and improve the embedding capacity. Compared with traditional compression methods, this method has a higher compression rate and there is no difference in semantic understanding between the simulated secret audio and the original secret audio.

Carrier audio is decomposed by two-level wavelet to obtain low-frequency, intermediate-frequency, and high-frequency sequences; then the intermediate-frequency coefficients are decomposed by first-level wavelet to obtain low-frequency and high-frequency sequences. Due to the robustness of the low-frequency sequence, but the low-frequency signal in the first wavelet decomposition represents the contour of the carrier audio, which cannot be changed, so the low-frequency signal in the second wavelet decomposition is selected as the hidden position, in terms of robustness and transparency. reach equilibrium.

The key is quantized and embedded into the carrier audio as a watermark, so that the voice hiding system does not need to open a secret channel to transmit the key. Due to the blind watermark hiding technology, the key can be extracted without the original carrier audio, which further illustrates the security of the system.

In the process of key watermark quantization and embedding, quantization is performed on the low-frequency coefficients of discrete cosine transform, which improves the robustness of the algorithm; in addition, the choice of quantization factor also controls the balance between robustness and transparency. In the key watermark extraction process, threshold quantization is used to further improve the robustness of the algorithm.

As shown in Figure 6, the present embodiment also provides an audio hiding device based on wavelet transform and quantization embedding key, the audio transform module is used to obtain carrier audio and secret audio, and carry out secondary wavelet transform to carrier audio to obtain the location of the hidden information to be embedded;

The extremum extraction module is used to extract the extremum and extremum coordinates of the secret audio, and obtain hidden information according to the extremum and extremum coordinates;

The audio embedding module is used to embed the concealed information into the carrier audio according to the position of the concealed information to be embedded to obtain the first audio;

The key embedding module is used to embed the key into the carrier audio that does not contain hidden information by using a quantization method to obtain the second audio, and concatenate the first audio and the second audio to obtain the transmission audio.

An audio hiding device based on wavelet transform and quantization embedding key in this embodiment can execute an audio hiding method based on wavelet transform and quantization embedding key provided in the method embodiment of the present invention, and can execute the method embodiment Any combination of implementation steps has the corresponding functions and beneficial effects of the method.

This embodiment also provides an audio hiding device based on wavelet transform and quantized embedded key, including:

at least one processor;

at least one memory for storing at least one program;

When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

An audio hiding device based on wavelet transform and quantization embedding key in this embodiment can execute an audio hiding method based on wavelet transform and quantization embedding key provided in the method embodiment of the present invention, and can execute the method embodiment Any combination of implementation steps has the corresponding functions and beneficial effects of the method.

It can be understood that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware and an appropriate combination thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit . Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). As known to those of ordinary skill in the art, the term computer storage media includes both volatile and nonvolatile media implemented in any method or technology for storage of information, such as computer readable instructions, data structures, program modules, or other data. permanent, removable and non-removable media. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cartridges, tape, magnetic disk storage or other magnetic storage devices, or can Any other medium used to store desired information and which can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media .

The embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those of ordinary skill in the art, various modifications can be made without departing from the gist of the present invention. kind of change.

Claims (8)

1. an audio concealment method based on wavelet transform and quantization embedding key, it is characterized in that, comprising the following steps: Obtain carrier audio and secret audio, perform secondary wavelet transform on the carrier audio, and obtain the position to be embedded with hidden information; Extract the extremum and extremum coordinates of the secret audio, and obtain hidden information according to the extremum and extremum coordinates; According to the location of the concealed information to be embedded, embed the concealed information into the carrier audio to obtain the first audio; use the quantization method to embed the key into the carrier audio that does not contain the concealed information to obtain the second audio, and concatenate the first audio and the second audio to obtain transmit audio; Said extracting the extremum and extremum coordinates of the secret audio, obtaining concealed information according to the extremum and extremum coordinates, include: extract the extrema and extrema coordinates of the secret audio; Extract the differential sequence of the extreme value coordinates, and perform stretching transformation on the extreme value and the difference sequence of the extreme value coordinates; perform Haar wavelet first-level reconstruction on the extremum and the difference sequence of the extreme value coordinates after the stretching transformation, to obtain hidden information; The second audio is obtained by using a quantization method to embed the key into the carrier audio that does not contain hidden information, Connecting the first audio and the second audio in series to obtain transmission audio, including: Obtain the length of the extremum sequence as the key, and convert the key into a binary code; performing discrete cosine transform on the carrier audio that does not contain concealed information to obtain a fourth low-frequency coefficient and a fourth high-frequency coefficient; Embedding the binary coded quantization into the fourth low-frequency coefficient to obtain the fifth low-frequency coefficient; Inverse discrete cosine transform is performed on the fifth low-frequency coefficient and the fourth high-frequency coefficient to obtain the carrier audio containing the key as the second audio; Connect the first audio and the second audio in series to obtain transmission audio. 2. a kind of audio hiding method based on wavelet transform and quantization embedding key according to claim 1, it is characterized in that, described carrier audio is carried out secondary wavelet transform, obtains the position to be embedded concealed information, comprises: Carrier audio is decomposed by two-level Haar wavelets to obtain the first high-frequency coefficient, the first intermediate-frequency coefficient and the first low-frequency coefficient; The Haar wavelet first-level decomposition is performed on the first intermediate-frequency coefficient to obtain the second high-frequency coefficient and the second low-frequency coefficient, and the position of the second low-frequency coefficient is used as the position to be embedded with hidden information. 3. a kind of audio concealment method based on wavelet transform and quantization embedding key according to claim 2, it is characterized in that, described according to the position to be embedded concealed information, concealed information is embedded in the carrier audio frequency, Get first audio, including: Embedding the hidden information into the second low-frequency coefficient to obtain the third low-frequency coefficient; The second high-frequency coefficient and the third low-frequency coefficient are subjected to the reconstruction of the second Haar wavelet transform to obtain the third intermediate-frequency coefficient; The first audio frequency is obtained by performing Haar wavelet first-level reconstruction on the first high-frequency coefficient, the third intermediate-frequency coefficient and the first low-frequency coefficient. 4. a kind of audio hiding method based on wavelet transform and quantization embedding key according to claim 1, It is characterized in that it also includes the following steps: Obtain the transmitted audio, and extract the key from the transmitted audio by discrete cosine transform; Combining secret key and quadratic wavelet transform to obtain secret audio from transmitted audio. 5. a kind of audio hiding method based on wavelet transform and quantization embedding key according to claim 4, It is characterized in that said adopting discrete cosine transform to extract key from transmission audio includes: Discrete cosine transform is performed on the transmission audio containing the key to obtain the sixth low-frequency coefficient; A binary code is obtained from the sixth low-frequency coefficient, and the binary code is converted into a key. 6. A kind of audio hiding method based on wavelet transform and quantization embedding key according to claim 5, it is characterized in that, described combination key and secondary wavelet transform obtain secret audio from transmission audio, comprising: Obtaining a third audio containing concealed information from the transmission audio according to the key, and performing Haar wavelet secondary decomposition on the third audio to obtain a fourth intermediate frequency coefficient; Perform Haar wavelet first-level decomposition on the fourth intermediate frequency coefficient to obtain the seventh low frequency coefficient; obtaining concealed information from the seventh low-frequency coefficient, the concealed information including extremums and differential sequences of extremum coordinates; Perform first-order spline interpolation on the extreme value and the difference sequence of the extreme value coordinates to obtain the secret audio. 7. An audio concealment device based on wavelet transform and quantization embedding key, characterized in that, comprising: The audio conversion module is used to obtain carrier audio and secret audio, and perform secondary wavelet transform on the carrier audio to obtain the position of the concealed information to be embedded; The extremum extraction module is used to extract the extremum and extremum coordinates of the secret audio, and obtain hidden information according to the extremum and extremum coordinates; The audio embedding module is used to embed the concealed information into the carrier audio according to the position of the concealed information to be embedded to obtain the first audio; The key embedding module is used to embed the key into the carrier audio that does not contain hidden information by using a quantization method, Obtaining the second audio, concatenating the first audio and the second audio to obtain the transmission audio; Said extracting the extremum and extremum coordinates of the secret audio, obtaining concealed information according to the extremum and extremum coordinates, include: extract the extrema and extrema coordinates of the secret audio; Extract the differential sequence of the extreme value coordinates, and perform stretching transformation on the extreme value and the difference sequence of the extreme value coordinates; perform Haar wavelet first-level reconstruction on the extremum and the difference sequence of the extreme value coordinates after the stretching transformation, to obtain hidden information; The second audio is obtained by using a quantization method to embed the key into the carrier audio that does not contain hidden information, Connecting the first audio and the second audio in series to obtain transmission audio, including: Obtain the length of the extremum sequence as the key, and convert the key into a binary code; performing discrete cosine transform on the carrier audio that does not contain concealed information to obtain a fourth low-frequency coefficient and a fourth high-frequency coefficient; Embedding the binary coded quantization into the fourth low-frequency coefficient to obtain the fifth low-frequency coefficient; Inverse discrete cosine transform is performed on the fifth low-frequency coefficient and the fourth high-frequency coefficient to obtain the carrier audio containing the key as the second audio; Connect the first audio and the second audio in series to obtain transmission audio. 8. An audio concealment device based on wavelet transform and quantization embedding key, characterized in that, comprising: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the audio concealment method based on wavelet transform and quantization embedding key according to any one of claims 1-6.