KR100966830B1 - Apparatus for inserting audio watermark, apparatus for detecting audio watermark and test automation system for detecting audio distortion using the same - Google Patents

Abstract

PURPOSE: An audio watermark inserting and detecting apparatus and an acoustic distortion detecting automation system using the same are provided to automatically test acoustic distortion without an inspector by testing the acoustic distortion using a digital watermarking technique and to determine even an acoustic distortion level. CONSTITUTION: An audio watermark inserting apparatus(310) is composed of a watermark inserting strength measuring unit, a watermark generator(311), a watermark PCM(pulse code modulation) generator, and a watermark inserting unit(314). The watermark inserting strength measuring unit measures the watermark inserting strength to be inserted into an original PCM data when the original PCM data is transformed into a DFT(discrete fourier transform) and is represented in the form of 2D power spectrum array. The watermark inserting unit inserts a watermark PCM data into the original PCM data.

Description

APPARATUS FOR INSERTING AUDIO WATERMARK, APPARATUS FOR DETECTING AUDIO WATERMARK AND TEST AUTOMATION SYSTEM FOR DETECTING AUDIO DISTORTION USING THE SAME}

Embodiments of the present invention relate to a technique capable of automatically detecting acoustic distortion.

In general, digital devices such as video, DVD players, digital television receivers, digital set-top boxes, digital multimedia broadcasting (DMB) phones, portable multimedia players (PMPs), MP 3 players, and the like, may have abnormalities after the manufacturing process is completed. You will test the quality of the video / audio.

Here, when testing the quality of audio, previously, the inspector manually listened to the audio signal output from the digital device for a long time (eg, 24 hours or more) and tested the quality of the audio.

In this case, the evaluation result is different according to various hearing environments affecting the sound, and the audio quality test takes a long time. In addition, since the audio test method using the inspector is evaluated and developed by a large number of groups, there is a problem in that development efficiency is not good because a lot of time, effort and cost are consumed.

Embodiments of the present invention are intended to exclude objectivity of the determination and ensure objectivity in the audio test by inserting a watermark into the audio signal and extracting the watermark from the audio signal to determine whether the sound is distorted.

Other technical problems by the embodiments of the present invention can be understood by the following description, which can be realized by the means and combinations thereof shown in the claims.

An audio watermark embedding apparatus according to an embodiment of the present invention, when the original Pulse Code Modulation (PCM) data is converted into a Discrete Fourier Transform (DFT) to represent a 2D Power Spectrum Array, A watermark embedding strength measuring device for measuring the embedding strength of the watermark to be inserted into the PCM data; A watermark generator configured to receive watermark embedding information and generate a watermark signal in the form of the 2D power spectrum array; A watermark PCM generator for generating watermark PCM data by applying an inverse discrete fourier transform (IDFT) after applying the measured watermark embedding intensity to the watermark signal; And a watermark inserter for inserting the watermark PCM data into the original PCM data.

The audio watermark detection apparatus according to an embodiment of the present invention, after receiving the PCM data in which the watermark in the form of a 2D Power Spectrum Array is inserted, and converts the PCM data into a Discrete Fourier Transform (DFT). A watermark signal predictor for converting and predicting the watermark signal; A synchronization signal generator for generating a synchronization signal for detecting the watermark using input synchronization information; A synchronization adjuster for adjusting synchronization for the watermark detection using the predicted watermark signal and the synchronization signal; And a watermark detector for detecting a watermark signal embedded in the PCM data.

According to an aspect of the present invention, there is provided an automatic system for detecting acoustic distortion using an audio watermark, including: a watermark insertion apparatus for generating watermark PCM data and inserting the watermark PCM data into input original PCM data; An audio acquisition unit including an audio output unit configured to receive the watermark-embedded PCM data, and output an audio output unit to a speaker, and an audio capture unit to capture the watermark-embedded PCM data output to the speaker; And a sound distortion determining apparatus for determining whether the sound is distorted by extracting a watermark signal from the watermark-embedded PCM data after receiving the captured watermark embedded PCM data.

According to another aspect of the present invention, there is provided an automatic system for detecting acoustic distortion using an audio watermark, including: a watermark embedding apparatus for generating watermark PCM data and inserting the watermark PCM data into input original PCM data; An audio output device configured to output the watermark-embedded PCM data to the outside after receiving the watermark-embedded PCM data; An audio acquisition device for acquiring the watermark-embedded PCM data output from the audio output device; And a sound distortion determination device for determining whether the sound is distorted by extracting a watermark signal from the watermark-embedded PCM data after receiving the obtained watermark-embedded PCM data.

According to the embodiments of the present invention, since the distortion of the sound is tested using a digital watermarking technology, the sound distortion test can be automatically performed without a separate examiner who tests the quality of the audio, so that the objectivity of the audio quality test is Can be secured.

Since the watermark is generated to reflect the strain caused by the signal distortion and then inserted into the audio signal, when the sound distortion occurs, it is possible to determine whether the sound distortion occurs and the degree of the sound distortion in detail.

In addition, it can automatically determine whether or not to determine the acoustic distortion, reducing the time, effort and cost of audio quality testing, and when applied to a real-time broadcast signal, reporting whether or not the acoustic distortion determination in real time to take quick action Can be taken.

Hereinafter, a specific embodiment of an audio watermark embedding apparatus and an audio watermark detection apparatus and an acoustic distortion determination automation system using the same will be described with reference to FIGS. 1 to 7. However, this is only an exemplary embodiment and the present invention is not limited thereto.

In describing the present invention, when it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. The following terms are defined in consideration of the functions of the present invention, and may be changed according to the intention or custom of the user, the operator, and the like. Therefore, the definition should be made based on the contents throughout the specification.

As a result, the technical spirit of the present invention is determined by the claims, and the following examples are one means for efficiently explaining the technical spirit of the present invention to those skilled in the art to which the present invention pertains. It is only.

In general, digital watermarking is concealed by inserting predetermined information in a signal form that cannot be visually or audibly recognized in multimedia content such as text, video, still images, and audio, and hiding hidden information. This technology is extracted and used as additional information for copyright, video certification and video monitoring.

In the embodiment of the present invention, the audio quality of the digital device is automatically tested using this digital watermarking technique. For example, if a watermark is inserted into original PCM data as an audio signal, and the PCM data with the watermark inserted therein is output from a predetermined digital device, the watermark is extracted and the watermark is extracted to determine whether acoustic distortion occurs. do. In this case, when a distortion occurs in the audio signal, the watermark is generated so that the degree of distortion due to the distortion can be well reflected.

Hereinafter, a technique of generating an audio watermark and inserting it into original PCM data and a technique of extracting a watermark from the PCM data into which the watermark is inserted will be described. .

1 is a diagram showing the configuration of an audio watermark embedding apparatus according to an embodiment of the present invention.

Referring to FIG. 1, the audio watermark embedding apparatus 100 may include a watermark embedding strength measuring unit 110, a watermark generator 120, a watermark pulse code modulation (PCM) generator 130, and a watermark inserter ( 140).

The watermark embedding strength measuring unit 110 receives original PCM (Pulse Code Modulation) data, and then converts the original PCM data into a Discrete Fourier Transform (DFT) to represent a 2D Power Spectrum Array. In this case, the insertion strength of the watermark to be inserted into the original PCM data is measured.

The watermark embedding strength measuring unit 110 may include a first partition unit 111, a first mask unit 113, a first discrete fourier transform (DFT) unit 115, and a watermark embedding strength calculating unit ( 117).

After receiving the original PCM data, the first partition unit 111 divides the original PCM data into frame units. For example, the first partition unit 111 divides the original PCM data sampled at 44,100 samples per second into frame units in which 1024 samples are grouped. In this case, a unit in which the divided frames are bundled by multiples of two (for example, two) is defined as a partition block.

As such, the reason for dividing the original PCM data by frame unit is that acoustic distortion (for example, distortion, mute, etc.) occurs in a short time (for example, 30 ms). Therefore, in order to cope with such sound distortion, the original PCM data is divided in units of frames, and a watermark is inserted in units of frames.

In addition, when using the partition block, it is possible to easily determine in which frame a distortion occurs by comparing the power spectral values between the frames belonging to each partition block.

The first mask unit 113 filters original PCM (Pulse Code Modulation) data divided in units of frames. In this case, negative distortion (distortion) occurring at the boundary of the watermark insertion period (for example, the boundary of the partition blocks) can be reduced. For example, a sinusoidal mask may be used as the first mask 113.

The first Discrete Fourier Transform (DFT) unit 115 performs discrete Fourier transform of the filtered original PCM data. In this case, the filtered original PCM data is divided into an amplitude component (ie, a power spectral component) and a phase component by a discrete Fourier transform. Here, the amplitude component represents the form of a 2D Power Spectrum Array by the Discrete Fourier Transform. In an embodiment of the present invention, a watermark is inserted into the amplitude component.

Specifically, the first DFT unit 115 represents each of the divided frames as a power spectrum component, and then divides the power spectrum of each frame non-linearly according to a frequency band.

For example, the first DFT unit 114 groups the power spectrum of each frame into two sample groups in the low frequency band, four sample groups in the middle frequency band, and eight sample groups in the high frequency band. Split nonlinearly according to frequency band. The power spectrum of each frame divided according to the frequency bands is sequentially arranged in the partition block unit.

 2 shows a 2D Power Spectrum Array. 2A is a graph illustrating a state in which a power spectrum of a partition block is nonlinearly divided according to frequency bands in units of partition blocks. FIG. 2B is a diagram illustrating the partition block of FIG. 2A in which one column corresponds to one partition block. Each partition block is composed of a plurality of blocks defined as partition bins according to frequency bands.

The watermark embedding strength calculator 117 calculates the embedding strength of the watermark when a watermark signal is inserted into the original PCM data in the form of the 2D power spectrum array.

Specifically, the watermark embedding strength calculator 117 sequentially obtains the first moment and the second moment with respect to the amplitude component of the original PCM data, and then applies a non-linear quantization function to the second moment. Get the watermark embedding strength by applying the function). At this time, the first moment of the amplitude component of the original PCM data can be obtained through the following equation (1).

Here, 1stMoment (i) represents the first moment of the amplitude component of the original PCM data, and Frame (j) represents the power spectral value of the j frame.

The second moment of the amplitude component of the original PCM data may be obtained through Equation 2 below.

Here, 2ndMoment (i) represents the second moment of the amplitude component of the original PCM data.

In addition, obtaining the watermark embedding strength by applying a nonlinear quantization function to the second moment is given by Equation 3 below.

이다.Where Q () represents a nonlinear quantization function and q represents the watermark embedding strength. And Lw (j) is

to be.

The watermark generator 120 receives watermark embedding information (for example, user information, copyright information, distribution stage information, etc.) and synchronization information, and then uses the watermark embedding information and synchronization information to generate the 2D power. Generate a watermark signal in the form of a spectral array.

The watermark generator 120 includes a message encoder 121, a pseudo random sequence generator 124, and a watermark signal generator 127.

After receiving the watermark embedding information, the message encoding unit 121 encodes the watermark embedding information using an error correction code such as a Reed-Solomon (RS) code or a Low Density Parity Check (LDPC) code. Message Encoding).

The reason why the watermark embedding information is encoded using an error correction code is to recover the original data even if the watermark embedding information is distorted or deformed.

The pseudo random sequence generator 124 generates a first pseudo random sequence using the message-encoded watermark embedding information as a seed value.

In addition, after receiving the synchronization information, the pseudo random sequence generator 124 generates a second pseudo random sequence using the synchronization information as a seed value. Here, the synchronization information refers to any information previously agreed for synchronization of the watermark in the audio watermark embedding apparatus and the audio watermark detection apparatus.

The synchronization information is periodically inserted in units of the partition block and includes time synchronization information for time axis synchronization of partition blocks, and block synchronization information for ordering partition blocks in the 2D power spectrum array.

Specifically, after the pseudo random sequence generator 124 receives the time synchronization information, generates the 2-1 pseudo random sequence using the time synchronization information as a seed value, and receives the block synchronization information. The 2-2 pseudo-random sequence is generated using the block synchronization information as a seed value.

The first pseudo random sequence and the second pseudo random sequence refer to a signal in which a value of 1 or -1 is randomly arranged according to the seed value.

The watermark signal generator 127 generates a watermark signal by configuring the first pseudorandom random sequence and the second pseudorandom random sequence in the form of a 2D power spectrum array.

In other words, the watermark signal generation unit 127 has the same shape as that of the 2D Power Spectrum Array according to the DFT transformation of the original PCM data with the first pseudo random sequence and the second pseudo random sequence. To generate a watermark signal. Here, the watermark signal is PCM data of an amplitude component.

Here, the first pseudo random sequence using the watermark embedding information is randomly inserted into the partition bins of partition blocks in the 2D power spectrum array, and the second pseudo random sequence using the synchronization information is inserted into a preset partition bin. . Here, the insertion position of the second pseudo random sequence is set through a random index function.

Specifically, in the 2-1 pseudo random sequence using the time synchronization information, the same pseudo random sequence is periodically inserted into a partition bin at a predetermined position in each partition block, and the 2-2 pseudo random sequence using the block synchronization information is used. The sequence is inserted with one pseudo-random sequence scattered into partition bins at predetermined positions in the 2D power spectral array.

Referring to FIG. 3, the 2D power spectrum array consists of five partition blocks, each partition block consisting of seven partition bins. In the 2D power spectrum array, an area in which a 2-1 like random sequence is inserted using time synchronization information and an area in which a 2-2 like random sequence is inserted using the block synchronization information may be arbitrarily distinguished.

For example, in the 2D power spectrum array, the first to third partition bins of each partition block are regions to which the 2-1 pseudo-random sequence is inserted, and the fourth to seventh partition bins of each partition block are the second-2. This is called a region into which a pseudorandom sequence is to be inserted.

Then, the same 2-1 pseudo-random sequence (T) is repeatedly inserted into the first to third partition bins in each partition block, and one first to the predetermined partition bin among the fourth to seventh partition bins in each partition block. 2-2 pseudo random sequence B is scattered and inserted.

A first pseudo random sequence using the watermark embedding information is randomly inserted into the partition bin in which the second-2 pseudo-random sequence B is not inserted among the fourth to seventh partition bins in each partition block. In this manner, the first pseudo random sequence and the second pseudo random sequence may be inserted to generate a watermark signal in the form of a 2D power spectrum array.

According to the embodiment of the present invention, the watermark signal is implemented in the form of a 2D power spectrum array in units of partition blocks, which are bundled in multiples of two, so that the insertion period of the watermark can be made in a very short time (for example, 30 ms). In this case, acoustic distortion occurring in a very short time can be detected accurately.

The watermark PCM generator 130 applies the measured watermark embedding strength to the watermark signal, and then converts an inverse discrete fourier transform (IDFT) to generate watermark PCM data.

The watermark PCM generator 130 includes a weight applying unit 131, an inverse discrete fourier transform (IDFT) unit 134, and a second mask unit 137.

The weight applier 131 applies the watermark embedding strength calculated by the weight calculator 117 to the watermark signal generated by the watermark signal generator 127.

In this case, when the first pseudo-random sequence and the second pseudo-random sequence value are 1, the weight applier 131 may, for example, power the n frames in a partition block composed of n frames and n + 1 frames. The spectral value is greater than the average power spectral value of the n frames and n + 1 frames, and the powermark embedding intensity is applied such that the power spectral value of the n + 1 frame is less than the average power spectral value. The n frames and n + 1 frames refer to frames arranged sequentially.

In addition, when the first pseudorandom random sequence and the second pseudorandom random sequence value are −1, the weight applying unit 131 may have a power spectrum value of the n frames and an average power spectrum of the n frames and n + 1 frames. And the powermark insertion strength is applied such that the power spectral value of the n + 1 frame is larger than the average power spectral value.

This is because the watermark detection apparatus to be described later predicts and detects a watermark by using a difference in power spectral values between n frames and n + 1 frames for each partition block constituting the 2D power spectrum array.

The IDFT (Inverse Discrete Fourier Transform) unit 134 converts the weighted watermark signal to inverse discrete Fourier transform. In this case, the IDFT unit 134 receives the phase component of the original PCM data from the first DFT unit 115 and then inverse discrete Fourier transform on the weighted watermark signal.

Since the weighted watermark signal is a signal composed only of an amplitude component, the IDFT unit 134 receives a phase component of the original PCM data from the first DFT unit 115 and applies it to the weighted watermark signal. Inverse Discrete Fourier Transform. In this case, the signal output from the IDFT unit 134 becomes a watermark signal in the form of PCM data having both an amplitude component and a phase component.

After receiving the watermark PCM data, the second mask unit 137 filters the watermark PCM data. Here, for example, a sinusoidal mask may be used as the second mask part 137. By filtering the watermark PCM data through the second mask unit 137, it is possible to reduce sound quality distortion generating elements.

The watermark inserter 140 inserts the watermark PCM data output from the second mask unit 137 into the original PCM data. Since the original PCM data and the watermark PCM data are signals of the same type, the watermark inserter 140 can insert the watermark PCM data into the original PCM data by superimposing the two signals. .

4 is a diagram illustrating a configuration of an audio watermark detection apparatus according to an embodiment of the present invention.

Referring to FIG. 4, the audio watermark detection apparatus 200 includes a watermark signal predictor 210, a synchronization signal generator 220, a synchronization adjuster 230, and a watermark detector 240.

The watermark signal predictor 210 receives PCM data into which a watermark in the form of a 2D Power Spectrum Array is inserted, and then converts the PCM data into a Discrete Fourier Transform (DFT) to convert the watermark signal. Predict.

The watermark signal predictor 210 includes a second partition unit 211, a third mask unit 213, a second DFT unit 215, and a watermark signal predictor 217.

After receiving the PCM data with the watermark embedded therein, the second partition unit 211 divides the PCM data with the watermark embedded into frames. In this case, the number of samplings forming the frame unit is the same as that of the first partition 111.

The third mask unit 213 filters the PCM data into which the watermark in the frame unit is inserted. In this case, acoustic distortion may be reduced at each boundary of the frame unit. For example, a sinusoidal mask may be used as the third mask part 213.

The second DFT unit 215 forms a 2D power spectrum array of amplitude components of the watermark-embedded PCM data by discrete Fourier transforming the watermark-embedded PCM data. In the embodiment of the present invention, since the watermark is embedded in the amplitude component of the PCM data, the watermark is also used as the amplitude component of the PCM data.

The watermark signal predictor 217 normalizes the amplitude of the watermark-embedded PCM data for each frequency band divided when the watermark is inserted, and then configures each partition block constituting the 2D power spectrum array. A watermark signal is predicted by using the difference of the normalized amplitudes between the frames belonging to.

In detail, the watermark signal predictor 217 normalizes the amplitude of the watermark-embedded PCM data for each frequency band divided when the watermark is inserted. In this case, the watermark signal predictor 217 may be normalized using Equation 4 below.

Where M (f) represents a normalized amplitude, N _band represents each frequency band, and M (f) represents an amplitude value in each frequency band.

Thereafter, the watermark signal predictor 217 obtains a difference of normalized amplitudes of frames belonging to the partition block for each partition block, in order to predict the watermark signal.

The reason for predicting the watermark signal through this is that, as described above, when the watermark embedding intensity is applied to the watermark signal, the difference in power spectral values between the frames belonging to the partition block depends on the value of the pseudo random sequence This is because the watermark embedding strength is applied to increase.

According to an embodiment of the present invention, by implementing the watermark signal in the form of a 2D power spectrum array in the unit of partition block grouped by a multiple of 2, by comparing the power spectral value between the two frames belonging to each partition block to the sound in any frame It is easy to determine whether distortion has occurred.

When acoustic distortion occurs in both frames belonging to the partition block, since the partition block itself has a different signal pattern from other partition blocks, it is easy to determine whether the corresponding partition block is acoustically distorted.

Here, in order to predict the watermark signal, a difference of a log scale of frames belonging to the partition block may be obtained for each partition block with respect to the normalized amplitude. In this case, the predicted watermark signal may be obtained through Equation 5 below.

Where | WM _i (f _j ) | represents the predicted watermark signal, | M _n (f _j ) | represents the normalized amplitude of the nth frame, and | M _{n + 1} (f _j ) | is n + Represents the normalized amplitude of the first frame. At this time, | M _n (f _j ) | and | M _{n + 1} (f _j ) | constitute one partition block.

When the log scale is used, when distortion occurs in the watermark signal due to compression of the watermark signal, analog-to-digital conversion, or the like, distortion of the distortion may be reduced. However, it is not necessary to use the log scale for the prediction of the watermark signal, and when the deformation is small, such as a clip of the watermark signal, the watermark signal can be predicted without using the log scale.

The watermark signal also includes a synchronization signal (ie, a time synchronization signal and a block synchronization signal). Accordingly, the watermark signal predictor 217 may predict the time synchronization signal and the block synchronization signal.

The synchronization signal generator 220 generates a synchronization signal using the input synchronization information. The synchronization information includes time synchronization information and block synchronization information. The synchronization signal generator 220 generates a reference time synchronization signal using the time synchronization information, and uses the block synchronization information to generate a reference block synchronization signal. Create The synchronization information is the same as the synchronization information used in the audio watermark embedding apparatus 100.

The synchronization adjuster 230 includes a time synchronizer 231 and a block synchronizer 234. The time synchronizer 231 adjusts time axis synchronization of the partition blocks by using the predicted time synchronization signal and the reference time synchronization signal. As a result, when the PCM data in which the watermark is inserted is divided into a clip-like form, the starting position at which the watermark is inserted may be detected.

The time synchronizer 231 measures similarity between the predicted time synchronization signal and the reference time synchronization signal, and adjusts time axis synchronization of the partition block by using a point having the maximum similarity. In this case, the similarity between the predicted time synchronization signal and the reference time synchronization signal may be obtained through Equation 6 below.

Here, C _timesync represents a similarity between the predicted time synchronization signal and the reference time synchronization signal, T _{sync (i)} represents a reference time synchronization signal, and E (T ' _{sync (i)} ) represents a predicted time synchronization signal. Indicates.

The block synchronizer 234 rearranges mutual positions (orders) between partition blocks in a 2D power spectrum array using the predicted block synchronization signal and the reference block synchronization signal.

For example, it is assumed that partition blocks are arranged in the order of partition block 1, partition block 2, and partition block 3 in the 2D power spectrum array when the watermark is inserted. However, if there is a change in position between partition blocks in the PCM data into which the watermark is inserted, the block synchronizer 234 rearranges the partition blocks in the original order.

The block synchronization unit 234 measures the similarity between the predicted block synchronization signal and the reference block synchronization signal, and measures the mutual position between the partition blocks using a point having the maximum similarity (ie, a block synchronization position). Arrange. In this case, the similarity between the predicted block synchronization signal and the reference block synchronization signal may be obtained through Equation 7 below.

Here, C _blocksync represents a similarity between the predicted block synchronization signal and the reference block synchronization signal, B _{sync (i)} represents a reference block synchronization signal, and E (B ' _{sync (i)} ) represents a predicted block synchronization signal. Indicates.

The watermark detector 240 includes a watermark detector 241 and a message decoder 244.

The watermark detector 241 detects a watermark signal from a 2D power spectrum array of the PCM data into which the watermark is embedded. After receiving the seed value, the watermark detector 241 measures the similarity between the generated watermark signal and the predicted watermark signal using the seed value, and uses the seedmark value having the maximum similarity value as the watermark signal. Extract with

For example, the watermark detection unit 241 receives 1 to 128 seed values one by one, and then measures the similarity between the generated watermark signal and the predicted watermark signal by using the received seed values. The seed value having the maximum similarity value is extracted as a watermark signal. Here, the audio watermark embedding apparatus 100 uses any one of the 1 to 128 seed values as the watermark embedding information.

Here, C _{codewore (i)} represents the similarity between the watermark signal generated using the seed value and the predicted watermark signal, and C _{seed (i)} represents the watermark signal generated using the seed value. , | WM _i (f _j ) | represents the predicted watermark signal.

The message decoding unit 244 corrects an error by applying an error correction algorithm to the detected watermark signal. That is, since the watermark signal is message encoded with an error correction code, the message decoding unit 244 detects an error through the error correction algorithm and then corrects the detected error.

Meanwhile, it is possible to detect whether the audio signal is distorted (such as sound distortion or cracking) by using an audio watermark embedding and extraction technique according to an embodiment of the present invention. Hereinafter, a description will be given of the automatic sound distortion detection system using the audio watermark.

5 is a diagram showing the configuration of the automatic system for detecting acoustic distortion according to an embodiment of the present invention.

Referring to FIG. 5, the acoustic distortion detection automation system 300 includes a watermark embedding apparatus 310, an audio acquisition apparatus 320, and an acoustic distortion determination apparatus 330.

The watermark embedding apparatus 310 generates watermark PCM data, and then inserts the watermark PCM data into the input original PCM data. The watermark embedding apparatus 310 includes a watermark generator 311 and a watermark inserter 314.

Since the watermark embedding apparatus 310 is the same as the configuration of the audio watermark embedding apparatus shown in FIG. 1, a detailed description thereof will be omitted. However, the watermark embedding apparatus 310 uses an audio perceptual distortion signal (ADPS) capable of detecting acoustic distortion as watermark embedding information.

Here, the watermark generator 311 is a component corresponding to the watermark embedding strength measuring instrument 110, the watermark generator 120, and the PCM watermark generator 130 in FIG. 1, and the watermark inserter 314 is a component corresponding to the watermark inserter 140 in FIG.

The original PCM data input to the watermark embedding apparatus 310 may be an audio signal stored in a file form, or may be a broadcast signal (eg, a digital broadcast signal or a radio broadcast signal) transmitted and input in real time.

That is, the watermark embedding apparatus 310 may insert the acoustic distortion detection signal as a watermark into an audio signal stored in a storage unit of a predetermined device, and insert the acoustic distortion detection signal into a watermark in a broadcast signal transmitted in real time. You may.

Here, when using a real-time broadcast signal, the watermark embedding apparatus 310 must be provided separately from the PCM data generation unit (not shown) for receiving the broadcast signal to generate real-time PCM data.

In this case, the PCM data generation unit (not shown) may store the generated PCM data in a file format in a storage unit (not shown), and the watermark embedding apparatus 310 may store a file in the storage unit (not shown). After extracting the PCM data stored in the form, a watermark may be inserted into the extracted PCM data.

The audio acquisition device 320 may be a device that receives PCM data including the watermark and outputs an audio signal. For example, the audio acquisition device 320 may be a digital television, an MP 3 player, a DVD player, or the like. The audio acquisition device 320 includes an audio output unit 321 and an audio capture unit 324.

The audio output unit 321 outputs the watermarked PCM data input to the audio acquisition device 320 to a speaker (not shown) provided in the audio acquisition device 320.

The audio capture unit 324 is connected to an output terminal of the audio output unit 321 to capture PCM data including watermarks output from the output terminal of the audio output unit 321 to the speaker (not shown). .

That is, the audio capture unit 324 is connected to the output terminal of the audio output unit 321 in the audio acquisition device 320 to capture the watermark embedded PCM data output to the speaker (not shown) .

In this case, since the audio capture unit 324 captures the PCM data embedded with the watermark output from the output terminal of the audio output unit 321, it is possible to capture a signal without external noise (Line-in system).

The sound distortion determining apparatus 330 receives the watermark embedded PCM data captured by the audio capture unit 324, and then extracts the watermark signal from the watermark embedded PCM data.

The acoustic distortion determination apparatus 330 determines whether acoustic distortion has occurred in the extracted watermark signal, and reports a result of the acoustic distortion. For example, when a watermark is inserted into a real-time broadcast signal, the acoustic distortion determination device 330 may report whether or not the acoustic distortion occurs in real time, and if an acoustic distortion occurs, an immediate action may be taken. .

6 is a view showing the configuration of the acoustic distortion determination apparatus according to an embodiment of the present invention.

Referring to FIG. 6, the acoustic distortion determination apparatus 330 includes a watermark detector 331, an acoustic distortion determination unit 332, a database 333, a reporting unit 334, and a display unit 335.

The watermark detector 331 detects a watermark signal from the watermarked PCM data. Since the watermark detector 331 is identical to the configuration of the audio watermark detection apparatus shown in FIG. 2, a detailed description thereof will be omitted. However, the watermark signal detected by the watermark detector 331 becomes the audio perceptual distortion signal (ADPS).

Since the watermark embedding apparatus 310 generates the watermark in the form of a 2D power spectrum array, the watermark period can be inserted with a very short watermark, thereby accurately detecting sound distortions occurring at a very short moment. . In addition, since it is possible to know in which frame among the frames belonging to the partition block, the acoustic distortion position can be accurately determined.

The acoustic distortion determination unit 332 measures the similarity between the watermark signal inserted by the watermark embedding apparatus 310 and the detected watermark signal, and if the similarity value does not exceed a preset threshold, acoustic distortion It is determined that this has occurred. In this case, the similarity between the inserted watermark signal and the detected watermark signal may be obtained by the following equation (9).

Here, Cnc represents the similarity between the inserted watermark signal and the detected watermark signal, APDS represents the inserted watermark signal, and E (APDS) represents the detected watermark signal.

Since the watermark signal is generated to reflect distortion caused by sound distortion, the sound distortion determination unit 332 may accurately determine whether sound distortion occurs and the degree of sound distortion.

The acoustic distortion determination unit 332 may set the threshold value in consideration of the characteristic and the length of the APDS signal, and adjust the precision of the acoustic distortion determination according to the threshold value. For example, if the length of the APDS signal is 100, setting the threshold value to 0.8 causes the error rate of the acoustic distortion determination to be about 10 ^-15 , and setting the threshold value to 0.45 causes the error rate of the acoustic distortion determination to be 10 ^-6. That's enough.

The database 333 stores the result of determining the acoustic distortion (for example, whether or not the acoustic distortion is generated), and stores the PCM data in which the acoustic distortion has occurred.

The reporting unit 334 generates a report for the automated sound quality test by using the determination result of the acoustic distortion determination unit 332. For example, the reporting unit 334 may create a list for each PCM data, such as whether or not the acoustic distortion occurs in the PCM data, and the position where the acoustic distortion occurs.

The display unit 335 displays the report for the automated sound quality test generated by the reporting unit 334 on the screen so that the user can check it.

7 is a diagram showing the configuration of the automatic system for detecting acoustic distortion according to another embodiment of the present invention.

Referring to FIG. 7, the acoustic distortion detection automation system 400 includes a watermark embedding apparatus 410, an audio output apparatus 420, an audio acquisition apparatus 430, and an acoustic distortion determination apparatus 440. Since the watermark embedding apparatus 410 and the sound distortion determining apparatus 440 are the same as those of FIG. 5, a detailed description thereof will be omitted.

The audio output device 420 receives the PCM data having the watermark embedded therein and outputs the same through a speaker (or a microphone).

The audio acquisition device 430 includes an audio acquisition unit 431 and a noise removing unit 434. The audio acquisition unit 431 acquires PCM data including the watermark output from the audio output device 420. The noise removing unit 434 removes external noise (or distortion) generated when acquiring the watermarked PCM data.

Since the audio acquisition device 430 acquires an audio signal output to an external space, external noise may be mixed with the obtained audio signal according to a surrounding environment, and distortion may occur in the acquired audio signal during the acquisition process. It may be.

Accordingly, the noise removing unit 434 removes external noise (or distortion) generated when the watermark is embedded with the PCM data.

Although the present invention has been described in detail with reference to exemplary embodiments above, those skilled in the art to which the present invention pertains can make various modifications to the above-described embodiments without departing from the scope of the present invention. Will understand.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined by the claims below and equivalents thereof.

1 is a diagram showing the configuration of an audio watermark embedding apparatus according to an embodiment of the present invention.

2 shows a 2D Power Spectrum Array.

3 is a view showing a state of generating a watermark signal in the form of a 2D power spectrum array according to an embodiment of the present invention.

4 is a diagram illustrating a configuration of an audio watermark detection apparatus according to an embodiment of the present invention.

5 is a diagram showing the configuration of the automatic sound distortion detection system according to an embodiment of the present invention.

6 is a view showing the configuration of an acoustic distortion determining apparatus according to an embodiment of the present invention.

7 is a view showing the configuration of the automatic sound distortion detection system according to another embodiment of the present invention.

Explanation of symbols on the main parts of the drawings

100: audio watermark embedding apparatus 110: watermark embedding strength measuring instrument

111: first partition portion 113: first mask portion

115: first DFT unit 117: weight calculation unit

120: watermark generator 121: message encoding unit

124: pseudo random sequence generator 127: watermark signal generator

130: watermark PCM generator 131: weight applying unit

134: IDFT portion 137: second mask portion

140: watermark inserter 200: audio watermark detection device

210: watermark signal predictor 211: second partition unit

213: third mask portion 215: second DFT portion

217: watermark signal predictor 220: synchronization signal generator

230: synchronization regulator 231: time synchronization unit

234: block synchronization unit 240: watermark detector

241: watermark detection unit 244: message decoding unit

300: automatic distortion detection detection system 310: watermark insertion device

311: watermark generator 314: watermark inserter

320: audio acquisition device 321: audio output unit

324: Audio capture unit 330: Sound distortion determination device

331: watermark detector 332: sound distortion determination unit

333: database 334: reporting unit

335 display unit 410: watermark insertion device

420: audio output device 430: audio acquisition device

431: Audio acquisition unit 434: Noise reduction unit

440: acoustic distortion determination device

Claims (34)

A watermark that measures the insertion strength of the watermark to be inserted into the original PCM data when original PCM (Pulse Code Modulation) data is converted into a Discrete Fourier Transform (DFT) to represent a 2D Power Spectrum Array. Insertion strength gauge; A watermark generator configured to receive watermark embedding information and generate a watermark signal in the form of the 2D power spectrum array; A watermark PCM generator for generating watermark PCM data by applying an inverse discrete fourier transform (IDFT) after applying the measured watermark embedding intensity to the watermark signal; And And a watermark inserter for inserting the watermark PCM data into the original PCM data. The method of claim 1, The watermark embedding strength measuring device, A partition unit dividing the original PCM data into frame units; A Discrete Fourier Transform (DFT) unit configured to DFT transform the divided frames to form the 2D power spectrum array; And And a watermark embedding strength calculating unit for calculating the embedding strength of the watermark. The method of claim 2, The DFT unit, And forming the 2D power spectrum array in a partition block unit in which the frame is divided into multiples of 2. Each partition block unit includes a plurality of partition bins. The method of claim 2, The watermark embedding strength measuring device, And a mask unit for filtering the original PCM data divided in the frame unit. The method of claim 2, The watermark embedding strength calculator, Equation

Where Q () is a nonlinear quantization function,

,

,

, Frame (j) is the power spectrum value of j frame) An audio watermark embedding apparatus for calculating the embedding strength of the watermark by using. The method of claim 3, The watermark generator, A message encoding unit for receiving the watermark embedding information and encoding the watermark embedding information using an error correction code; A pseudo random sequence generator for generating a first pseudo random sequence using the message-encoded watermark embedding information as a seed value and a second pseudo random sequence using the input synchronization information as a seed value; And And a watermark signal generator configured to generate the watermark signal by configuring the first pseudorandom random sequence and the second pseudorandom random sequence in the form of the 2D power spectrum array. The method of claim 6, The synchronization information includes time synchronization information and block synchronization information. The pseudo random sequence generation unit generates a 2-1 pseudo random sequence using the time synchronization information as a seed value and generates a 2-2 pseudo random sequence using the block synchronization information as a seed value. Insert device. The method of claim 7, wherein The watermark signal generator, The first pseudo-random sequence is randomly inserted into a partition bin of the 2D power spectrum array, and the second pseudo-random sequence is inserted into a predetermined position by a random index function of the partition bin of the 2D power spectrum array. Watermark insertion device. The method of claim 6, The watermark PCM generator, A weight applying unit which applies the watermark embedding strength to the watermark signal; And And an inverse discrete fourier transform (IDFT) unit for generating a watermark signal in the form of PCM data by IDFT converting the watermark signal to which the watermark embedding strength is applied. 10. The method of claim 9, The partition block consists of two frames, The weight applying unit, When the first pseudorandom random sequence and the second pseudorandom random sequence have a value of 1, a power spectral value of a previous frame among two frames belonging to the corresponding partition block in each partition block of the 2D power spectrum array is assigned to the partition block. An audio watermark embedding, wherein the watermark embedding intensity is applied to be greater than an average power spectral value of the belonging frames, and so that the power spectral value of a frame of the second of the two frames belonging to the partition block is less than the average power spectral value. Device. The method of claim 10, The weight applying unit, When the value of the first pseudorandom random sequence and the second pseudorandom random sequence is −1, the power spectral value of the frame of the preceding frame among the two frames belonging to the corresponding partition block in each partition block of the 2D power spectrum array is the partition block. And applying the watermark embedding intensity such that a power spectral value of a frame next to the second of the two frames belonging to the partition block is larger than the average power spectral value. Insert device. After receiving the PCM data in which the watermark in the form of 2D Power Spectrum Array is inserted, the PCM data is transformed by DFT (Discrete Fourier Transform), and the inserted watermark signal is converted from the converted PCM data. Predicting watermark signal predictor; A synchronization signal generator for generating a synchronization signal for detecting the watermark using input synchronization information; A synchronization adjuster for adjusting synchronization for the watermark detection using the predicted watermark signal and the synchronization signal; And And a watermark detector for detecting a watermark signal embedded in the PCM data. The method of claim 12, The watermark signal predictor, A partition unit for dividing the watermark-embedded PCM data into frame units; A Discrete Fourier Transform (DFT) unit configured to perform a DFT transform on each of the divided frames to form the 2D power spectral array composed of partition blocks uniting the frames in multiples of two; And And a watermark signal predictor for normalizing the amplitude of each divided frame for each frequency band and predicting the watermark signal using a difference between normalized amplitudes of frames belonging to the partition block. . The method of claim 13, The watermark signal prediction unit, Equation

Where WM _i (f _j ) is the predicted watermark signal, M _n (f _j ) is the normalized amplitude of the n th frame, and M _{n + 1} (f _j ) is the n + 1 th Normalized amplitude of the frame) Predicting the watermark signal by using a. The method of claim 14, The watermark includes time synchronization information and block synchronization information. And the watermark signal predictor predicts a time synchronization signal and a block synchronization signal of the watermark. The method of claim 15, The synchronization signal generator, And generating a reference time synchronization signal using the input time synchronization signal, and generating a reference block synchronization signal using the input block synchronization information. The method of claim 16, The synchronization regulator, A time synchronizer for adjusting time axis synchronization of partition blocks of the predicted watermark using the predicted time synchronization signal and the reference time synchronization signal; And And a block synchronizer configured to adjust positions of partition blocks of the predicted watermark using the predicted block synchronization signal and the reference block synchronization signal. The method of claim 14, The watermark detector, And measuring a similarity between the generated watermark signal and the predicted watermark signal using a predetermined seed value, and extracting a seed value having the maximum similarity as a watermark signal. The method of claim 18, The watermark detector, Equation

Where C _{codewore (i)} is the similarity between the watermark signal generated using the seed value and the predicted watermark signal, C _{seed (i)} is the watermark signal generated using the seed value, WM _i (f _j ) | is the predicted watermark signal) And measuring the similarity between the generated watermark signal and the predicted watermark signal using the seed value. A watermark embedding apparatus for generating watermark PCM data and inserting the watermark PCM data into input original PCM data; An audio acquisition unit including an audio output unit configured to receive the watermark-embedded PCM data, and output an audio output unit to a speaker, and an audio capture unit to capture the watermark-embedded PCM data output to the speaker; And And receiving a watermark-embedded PCM data, and extracting a watermark signal from the watermark-embedded PCM data to determine whether the sound is distorted. The watermark embedding apparatus, when the original Pulse Code Modulation (PCM) data is converted into a Discrete Fourier Transform (DFT) to represent a 2D Power Spectrum Array, the watermark to be inserted into the original PCM data Watermark embedding strength measuring device for measuring the insertion strength of the watermark generator to receive the watermark embedding information to generate the watermark signal of the 2D power spectrum array type, applying the measured watermark embedding strength to the watermark signal And a watermark PCM generator for generating watermark PCM data by Inverse Discrete Fourier Transform (IDFT) conversion, and a watermark inserter for inserting the watermark PCM data into the original PCM data. Distortion Detection Automation System. 21. The method of claim 20, The watermark embedding apparatus, And a PCM data generation unit for receiving real-time broadcast signals and generating real-time original PCM data. The method of claim 21, The watermark embedding apparatus, Further comprising a storage unit for storing the generated original PCM data, And extracting the original PCM data from the storage unit and inserting the watermark PCM data into the extracted original PCM data. delete 21. The method of claim 20, The watermark generator, The audio distortion detection automation system using an audio watermark, using an audio perceptual distortion signal (ADPS) as the watermark embedding information. The method of claim 24, The sound distortion determination device, A watermark detector for detecting the watermark signal in the watermarked PCM data; An acoustic distortion determination unit that determines whether the acoustic distortion is determined by measuring a similarity between the watermark signal inserted by the watermark embedding apparatus and the detected watermark signal; And And a reporting unit for generating a report for automated sound quality test using the result of the acoustic distortion determination unit. The method of claim 25, The sound distortion determination unit, Equation

(Cnc is the similarity between the inserted watermark signal and the detected watermark signal, APDS is the inserted watermark signal, and E (APDS) is the detected watermark signal) And measuring the similarity between the watermark signal inserted by the watermark embedding apparatus and the detected watermark signal by using an audio watermark. The method of claim 25, The watermark detector, After receiving the PCM data in which the watermark in the form of 2D Power Spectrum Array is inserted, the PCM data is transformed by DFT (Discrete Fourier Transform), and the inserted watermark signal is converted from the converted PCM data. A watermark signal predictor for predicting; A synchronization signal generator for generating a synchronization signal for detecting the watermark using input synchronization information; A synchronization adjusting unit for adjusting synchronization for the watermark detection using the predicted watermark signal and the synchronization signal; And And a watermark detector for detecting a watermark signal embedded in the PCM data. A watermark embedding apparatus for generating watermark PCM data and inserting the watermark PCM data into input original PCM data; An audio output device configured to output the watermark-embedded PCM data to the outside after receiving the watermark-embedded PCM data; An audio acquisition device for acquiring the watermark-embedded PCM data output from the audio output device; And And receiving a watermark-embedded PCM data, and extracting a watermark signal from the watermark-embedded PCM data to determine whether the sound is distorted. The watermark embedding apparatus, when the original Pulse Code Modulation (PCM) data is converted into a Discrete Fourier Transform (DFT) to represent a 2D Power Spectrum Array, the watermark to be inserted into the original PCM data Watermark embedding strength measuring device for measuring the insertion strength of the watermark generator to receive the watermark embedding information to generate the watermark signal of the 2D power spectrum array type, applying the measured watermark embedding strength to the watermark signal And a watermark PCM generator for generating watermark PCM data by Inverse Discrete Fourier Transform (IDFT) conversion, and a watermark inserter for inserting the watermark PCM data into the original PCM data. Distortion Detection Automation System. The method of claim 28, The watermark embedding apparatus, And a PCM data generation unit for receiving real-time broadcast signals and generating real-time original PCM data. delete The method of claim 28, The watermark generator, The audio distortion detection automation system using an audio watermark, using an audio perceptual distortion signal (ADPS) as the watermark embedding information. The method of claim 28, The audio acquisition device, An audio acquisition unit for acquiring PCM data including the watermark output from the audio output device; And And a noise removing unit for removing external noise of the acquired watermark embedded PCM data. The method of claim 31, wherein The sound distortion determination device, A watermark detector for detecting the watermark signal in the watermarked PCM data; An acoustic distortion determination unit that determines whether the acoustic distortion is determined by measuring a similarity between the watermark signal inserted by the watermark embedding apparatus and the detected watermark signal; And And a reporting unit for generating a report for automated sound quality test using the result of the acoustic distortion determination unit. 34. The method of claim 33, The watermark detector, After receiving the PCM data in which the watermark in the form of 2D Power Spectrum Array is inserted, the PCM data is transformed by DFT (Discrete Fourier Transform), and the inserted watermark signal is converted from the converted PCM data. A watermark signal predictor for predicting; A synchronization signal generator for generating a synchronization signal for detecting the watermark using input synchronization information; A synchronization adjusting unit for adjusting synchronization for the watermark detection using the predicted watermark signal and the synchronization signal; And And a watermark detector for detecting a watermark signal embedded in the PCM data.

Images