JP2006171110A - Method for embedding additional information to audio data, method for reading embedded additional information from audio data, and apparatus therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for embedding additional information to audio data, corresponding to analog and digital transmission and corresponding to monaural transmission and irreversible compression, and to provide a method for reading the embedded additional information from the audio data and apparatuses therefor. <P>SOLUTION: Digitized discrete amplitude signals are transformed from a time region to a frequency region and the obtained spectra are divided to two groups in units of odd numbers, even numbers, etc. By adjusting gains, such that the summation of the amplitude of the one group is "1" if greater than the summation of the amplitude of the other and is "0" in the opposite case, the data are superposed and is returned to the original time region by inverse transformation. The data superposed signal are retransformed to the frequency region, and the comparison of the amplitude sum of the spectra of the two groups is performed. The embedded data are then taken out. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a method for embedding additional information in audio data, a method for embedding additional information from audio data, and devices thereof, and more particularly, a method for embedding additional information in audio data capable of embedding additional signal data in audio data, The present invention relates to a method and apparatus for embedding additional information from audio data.

  Currently, digitized music content is provided via the Internet in addition to CDs. With such digital audio data, there is no deterioration in sound quality no matter how many times it is copied. For this reason, illegal copy prevention technology is important.

  In order to prevent these unauthorized duplications, it is necessary to clarify the whereabouts of copyright by embedding copyright information etc. in the audio data itself in advance, and illegal copies are circulated by embedding content distribution destination information. It is possible to track the outflow route even in the case of an accident. At this time, in order not to lower the product value of the content, the change in sound quality due to the embedding of information must not be discriminated by the listener.

  In addition, music content is subjected to processing such as filtering, compression / decompression such as MPEG, AC3, and ATRAC, conversion from digital to analog, conversion from analog to digital, extraction, and change in playback speed. The embedded information needs to be stored for data value changes, loss, insertion, and resampling within a range that does not cause significant deterioration in sound quality.

  In view of such a current situation, a digital watermark technique has been proposed in which other data is superimposed on a sound signal so that it is difficult for humans to hear. In digital watermarking, in addition to basic methods such as adding weak pseudo-random signals directly to the original sound using spread spectrum, a psychoacoustic model that mimics hearing is used to avoid damaging the quality of the original sound as much as possible. Many use it to embed data without compromising sound quality.

  Typical psychoacoustic models used for data embedding include minimum audible limit, frequency masking (simultaneous masking), time masking, band elimination, and insensitivity to phase change.

  Patent Document 1 describes an example using the minimum audible limit. This technology uses the fact that hearing is sensitive around 3 to 4 kHz and is extremely insensitive to high and low frequency weak sounds, and by suppressing the frequency characteristics of data added to the original sound below the audible limit, Suppresses changes in sound quality due to data embedding.

  Patent Document 2 describes an example using frequency masking. In this technique, when there is a large frequency component in the adjacent frequency, the small component cannot be heard and is embedded using a phenomenon.

  Patent Document 3 describes an example using time masking. This technique utilizes the fact that even if a small sound is played immediately after a loud sound, it cannot be heard, and is called an echo method, an echo diffusion method, or the like.

  Patent Document 4 describes an example using band elimination. This utilizes the fact that it is difficult to hear even if a frequency band with no signal is created in a part of the audible band, and attempts to propagate in the air with or without the band.

  The psychoacoustic models listed above are similarly used in the irreversible compression technique of sound. That is, in the irreversible compression of the sound that is formed by reducing the reproducibility of the signal that is difficult to hear, the psychoacoustic model is used, and the allocation of information to the component that is difficult to hear is reduced as the compression is increased. For this reason, most of them are vulnerable to lossy compression. In particular, the method of embedding under the minimum audible limit and the method of creating an area without sound such as band elimination are very vulnerable to irreversible compression because the area is first targeted for information reduction.

  Patent Document 5 describes an example using phase change. This is a technique for embedding a watermark by applying phase modulation with a data signal by utilizing the fact that hearing is insensitive to phase changes. However, since a sound with a carrier signal to be used as a phase reference is required, it requires an original sound with only a carrier signal added by phase modulation, or other channel sound, and it must be compared with the original sound or another channel. Cannot extract data, and it is particularly difficult to extract data only from a sound in which data is embedded in a monaural signal.

  Many of the above methods often require the original sound itself or a key that is a parameter used at the time of embedding in order to distinguish the embedded data from the original sound. Although there are methods such as changing the amplitude directly and altering the compressed data, they have similar problems. For this reason, it is not suitable for use in the purpose of discriminating the target sound itself.

In many of these systems, it is assumed that data is extracted by calculating a digitized signal, and therefore many are not assumed to be temporarily analogized.
Patent Document 6 describes a method using movement of sound field localization. This embeds a multi-channel sound signal by changing the volume between channels and shifting the sound localization position, but cannot be used for a monaural signal. Propagation in the air is also impossible due to sound mixing in the air. The same applies to the first phase modulation.

Special table 2003-536113 gazette JP 2003-263183 A JP 2000-172282 A JP 2001-148670 A JP 2004-279469 A

  In the above prior art, those using a general psychoacoustic model are weak against lossy compression, do not consider analogization at the time of transmission only for digital signals, require original sound or key to extract data, monaural There were problems such as not supporting the signal. In addition, it is necessary to be able to extract data from analog audio to support analog broadcasting and analog communication channels, and it is necessary to extract data when the target sound is unknown, so no original sound or keys are required Therefore, it is required to withstand irreversible compression / decompression processing during transmission of sound signals and to be applicable to air propagation.

  SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a method for embedding additional information in audio data corresponding to analog and digital transmission, and corresponding to monaural transmission and irreversible compression, a method for embedding additional information from audio data, and an apparatus thereof. And

  The invention of claim 1 is a method for embedding additional information in audio data, wherein the additional information signal is superimposed on the frequency amplitude characteristic of the spectral distribution of the audio data.

  According to a second aspect of the present invention, in the additional information embedding method in the voice data of the first aspect, a discrete amplitude signal in the time domain of the digitized voice data is converted into the frequency domain, and the frequency sequence is converted into the converted signal. Additional information is superimposed on the data, and the frequency series data on which the additional data is superimposed is converted into audio data.

  According to a third aspect of the present invention, in the additional information embedding method in the voice data according to the first or second aspect, the spectrum distribution of the voice data is extracted, and a predetermined frequency distribution area of the spectrum distribution is divided into a plurality of bands. The divided bands are grouped into a plurality of data groups, the gain of the plurality of data groups is increased or decreased based on the additional information value, and the plurality of data groups after embedding the re-addition information is organized into a continuous spectrum distribution. And generating voice data to which additional information is added.

  According to a fourth aspect of the present invention, in the additional information embedding method according to any one of the first to third aspects, the number of groups of data groups to be collected is two, and 1-bit information is obtained depending on the total amplitude of the data groups. It is characterized by assigning.

  According to a fifth aspect of the present invention, in the method for embedding additional information in the audio data according to any one of the first to fourth aspects, the preprocessing is performed so that the average value of the total amplitude of the two data groups is substantially equal. It is characterized by.

  According to a sixth aspect of the present invention, in the method for embedding additional information in the audio data according to any one of the first to fifth aspects, conversion from time series data to frequency series data is performed by FFT (fast Fourier transform), DFT (discrete Fourier transform). Conversion), DST (discrete sine transform), DCT (discrete cosine transform), and MDCT (modified discrete cosine transform).

  According to a seventh aspect of the present invention, in the method for embedding additional information in the voice data according to any one of the first to sixth aspects, a plurality of voice data areas are selected, processing is performed for each selected area, and the number of the selected plurality of areas is set. Corresponding additional information of a plurality of bits is embedded.

  According to an eighth aspect of the present invention, in the additional information embedding method according to any one of the first to sixth aspects, the frequency range in which the additional information is embedded avoids a band in which the reproduced sound has a large influence on the voice recognition of the listener. It is characterized by performing.

  According to a ninth aspect of the present invention, in the additional information embedding method according to any one of the first to eighth aspects, the adjacent frames in the time domain are gently overlapped to relate to the frequency sequence data. Features.

  The invention of claim 10 extracts a spectral distribution of audio data, divides a predetermined frequency distribution region of the spectral distribution into a plurality of bands, collects the divided bands into a plurality of data groups, and This is a method for extracting additional information from audio data, in which the amplitude of the group is determined for each data group, and additional information is extracted based on the size relationship of each data group.

  According to a first aspect of the present invention, in the method for reading additional information from the voice data according to the tenth aspect, the division and summarization of the voice frequency is performed in synchronization with the additional signal.

  According to a twelfth aspect of the present invention, there is provided an apparatus for embedding additional information in audio data, comprising amplitude characteristic extracting means for extracting frequency amplitude characteristics of a spectral distribution of audio data, and superimposing means for superimposing additional information signals on the frequency amplitude characteristics. It is.

  According to a thirteenth aspect of the present invention, in the additional information embedding device in the voice data of the twelfth aspect, the additional information superimposing means converts the discrete amplitude signal in the time domain of the digitized voice data into the frequency domain and converts the signal. And a conversion means for superimposing additional information on the frequency sequence data on the received signal and converting the frequency sequence data on which the additional data is superimposed into audio data.

  According to a fourteenth aspect of the present invention, in the additional information embedding device according to the twelfth or thirteenth aspect, a spectral distribution extracting means for extracting a spectral distribution of the voice data, and a plurality of predetermined frequency distribution areas of the spectral distribution are arranged. Band division means for dividing the data into bands, band division means for collecting the divided bands into a plurality of data groups, and gains of the plurality of data groups are increased or decreased based on the additional information value, and the plurality of data after the re-addition information is embedded Additional information adding means for organizing the group into a continuous spectrum distribution and voice data creating means for creating voice data to which additional information is added based on the spectrum distribution are provided.

  According to a fifteenth aspect of the present invention, in the additional information embedding device according to any one of the twelfth to fourteenth aspects, the band dividing unit sets the number of groups of data groups to be collected to be two, and the additional information adding unit is the data One bit of information is assigned according to the total amplitude of the group.

  According to a sixteenth aspect of the present invention, there is provided the additional information embedding device according to any one of the twelfth to fifteenth aspects, further comprising preprocessing means for making the average value of the total amplitude of the two data groups substantially equal. It is characterized by that.

  According to a seventeenth aspect of the present invention, in the additional information embedding method according to any one of the twelfth to sixteenth aspects, the amplitude characteristic extracting means converts the time series data into the frequency series data by using FFT (Fast Fourier Transform). ), DFT (Discrete Fourier Transform), DST (Discrete Sine Transform), DCT (Discrete Cosine Transform), and MDCT (Modified Discrete Cosine Transform).

  The invention according to claim 18 is the additional information embedding device for audio data according to claims 12 to 17, wherein a plurality of frequency regions of the audio data are selected, the selected regions are processed, and the number of selected regions is selected. The additional information of a plurality of bits corresponding to is embedded.

  According to a nineteenth aspect of the present invention, in the additional information embedding device according to any one of the twelfth to eighteenth aspects, the additional information superimposing means causes the reproduction sound to affect the listener's voice recognition in the frequency region in which the additional information is embedded. It is characterized by avoiding a large band.

  According to a twentieth aspect of the present invention, in the additional information embedding method according to any one of the twelfth to nineteenth aspects, the amplitude characteristic extracting unit gently frames adjacent frames in the time domain when converting to the frequency series data. It is characterized by being overlapped and related.

  The invention according to claim 21 is a spectrum distribution extracting means for extracting a spectrum distribution of audio data, a band dividing means for dividing a predetermined frequency distribution area of the spectrum distribution into a plurality of bands, and a plurality of the divided bands. The additional information to the audio data is provided with an additional information extracting unit that collects the data groups, determines the magnitudes of the plurality of data groups for each data group, and extracts additional information based on the magnitude relationship of the data groups. It is a take-out device.

  According to a twenty-second aspect of the present invention, in the additional information reading apparatus for the voice data according to the twentieth aspect, the band dividing unit includes a frame synchronization unit that tunes to the audio frequency component and the combined additional signal.

  According to the present invention, additional data is superposed on audio data, and superposition and extraction of the additional data correspond to analog and digital transmission, and can correspond to monaural transmission and irreversible compression. This makes it possible to determine what is being broadcast on TV, radio, etc., and what is recorded on CDs and DVDs. More detailed information can be obtained from other devices connected to the Internet, for example. Operations such as obtaining from a computer or presenting the data itself can be performed.

  Furthermore, by superimposing character information such as subtitles and other audio data on the sound signal, the accessibility of the hearing impaired is improved, the presentation of various information at the time of disaster, etc., explanation using sub-speech and other languages This information can be inserted. In addition, by changing the data to be superimposed even with sounds that can be heard equally, it is possible to apply to the security field such as authentication by sound.

  Hereinafter, the best mode for carrying out the present invention will be described with reference to the drawings.

  First, an additional information embedding device for audio data according to this example will be described. FIG. 1 is a block diagram showing a device for embedding additional information in audio data according to an embodiment of the present invention, and FIG. 2 is a flowchart showing a method for embedding additional information in audio data according to an embodiment of the present invention.

  The apparatus for embedding additional information in audio data according to this example embeds an additional signal, which is a “watermark” that identifies the source in the audio signal. In this example, as shown in FIG. 1, the additional information embedding device into the audio data includes an audio signal storage means 11 for storing the audio data to be embedded with the additional signal as analog data, an A / D conversion means 12, An FFT converter 13 that converts a digitized discrete amplitude signal from the time domain to the frequency domain and divides the signal into two groups; and a signal averaging means 14 that equalizes the sum of the spectrum amplitudes of the two divided groups; Additional signal embedding means (watermark signal embedding means) 15 is provided. The additional signal embedding means 15 adjusts the gain so that the sum of the amplitudes of one group is “1” when the sum of the amplitudes of the other group is larger than the sum of the amplitudes of the other group, and “0” in the opposite case. Superimpose. In this example, the additional information embedding device includes an additional signal storage unit 16 that stores a watermark signal to be added to an audio signal, an inverse FFT conversion unit 17 that converts a discrete amplitude signal from a frequency domain to a time domain, It comprises a DA converter 18 for converting a digital signal into an analog signal.

  In this example, the gain of the portion where the spectrum of the original sound exists is adjusted, and the frequency characteristics are given fine strength, that is, unevenness as shown in FIG. The unevenness of the frequency characteristic is unlikely to be perceived by humans, unlike the resonant one. Actually, when the same sound is played from two or more speakers at the same time, if the distance is different, unevenness due to interference occurs, but there is almost no sense of incongruity in hearing. Then it can be said that it is obvious.

  The operation of this example will be described below with reference to the flowchart shown in FIG. 2 and FIGS. In this example, the analog audio signal read from the audio signal storage unit 11 is converted into digital data by the A / D conversion unit 12 as shown in FIG. 6 (ST1). In the example shown in the figure, for one frame, for example, 32 discrete values, the FFT transforming means 13 extracts the spectrum distribution of the digital audio data by FFT transform, and a predetermined frequency distribution region of the spectrum distribution is a frequency. Are alternately divided into two bands (odd order (O) and even order (E)) (ST3). Such frames are selected and processed for the required number of data lengths.

  As a result, the time-series analog data waveform of FIG. 7A is converted into an odd-order (O) and even-order (E) digital signal between predetermined frequencies along the frequency train as shown in FIG. 7B. Two data groups can be grouped. The number of discrete values is arbitrary, and can be, for example, 256 or more. As a result, the amplitude signal is converted from the time domain to the frequency domain.

  Next, in this example, the signal averaging means 14 performs preprocessing to make the average values of the total amplitudes of the two data groups substantially equal (ST4). That is, the gain of the plurality of data groups is increased or decreased based on the additional information value, and the plurality of data groups after the re-addition information is embedded is organized into a continuous spectrum distribution.

As shown in FIG. 8, processing is performed so that each coefficient is multiplied so that Σ (E) = Σ (O). However, the number of points in one frame is large (eg, 1024 points or more). In the case of musical tones with various sounds, statistically ΣE≈ΣO, so it can often be omitted. If it is embedded only in a specific frequency range, it is desirable to change the coefficient between the embedded part and the non-embedded part.

  In this example, watermark signal embedding processing is performed (ST5). As shown in FIG. 9, when the sum of the amplitudes of one group is larger than the sum of the amplitudes of the other group among the two groups in the odd-order (O) and even-order (E) units, Each next gain is adjusted so that it is “1” and “0” in the opposite case. Thereby, 1-bit information can be added between predetermined frequencies.

  In other words, the gains of the odd-order (O) and even-order (E) spectra are manipulated so that ΣE> ΣO if the embedded data is 1, and ΣE <ΣO if 0. In this example, the necessary number of frames is selected for the number of additional information bits to be embedded, and this calculation is performed for each frame, thereby embedding one bit in one frame and embedding data of the necessary number of bits. . At this time, it is preferable to target a specific frequency component. That is, it is preferable to avoid low-frequency, high-frequency, and auditory sensitive areas.

  As shown in FIG. 10, the data to which the additional signal is attached is inversely transformed by the inverse FFT transforming means 17 (ST7), returned to the original time domain, and restored to an audio signal by D / A conversion. Through these processes, the additional signal is embedded and output in the whole spectrum or a part of the spectrum (ST8).

  FIG. 12 shows the time series data (upper stage) and frequency series data (lower stage) of the audio signal before embedding the additional signal, and FIG. 13 shows the time series data (upper stage) and frequency series data (lower stage) of the audio signal after the additional signal is embedded. Indicates. Comparing both figures, it can be seen that there is almost no difference in time series data and there is almost no change in hearing. It can also be seen that the frequency series data after embedding the additional signal has irregularities due to the additional signal. In this example, there is no signal above 8 kHz, and it is necessary to add a noise component to embed an additional signal in a band above 8 kHz.

  In this way, the watermark signal is embedded, and even if there is unevenness in the spectrum, it is a change of only a few dB, and it is erased by compression, as when using band elimination, absolute audible limit, frequency and time masking. It will never be. In the present invention, even if the spectrum whose gain has been reduced disappears on the same principle as frequency masking, the spectrum whose gain has been increased remains and the amplitude difference is widened. is there.

  Next, an additional information device from audio data will be described. FIG. 3 is a block diagram showing a device for extracting additional information from audio data according to the embodiment of the present invention, and FIG. 4 is a flowchart showing a method for extracting additional information from audio data according to the embodiment of the present invention.

  In this example, as shown in FIG. 3, the device for extracting additional information from audio data includes a watermarked audio signal storage means 21 that stores the audio data in which the additional signal is embedded as analog data, and D / A conversion. Means 22 and frame synchronization for converting the digitized discrete amplitude signal from the time domain to the frequency domain to synchronize the FFT converter 24 and the FFT converter 24 so that the conversion is performed in the frame in which the additional signal is embedded. It comprises means 23 and additional signal extraction means 25 for comparing the amplitude sums of the two groups of spectra which are FFT-transformed signals and extracting watermark data.

  In this example, the frame synchronization means 23 eliminates the difference between the frame in which the additional signal is embedded and the frame being processed. Therefore, the frame being processed is moved, and ΣE and ΣO are compared. As shown in FIG. 11, the frame being processed matches the frame in which the additional signal is embedded at the point where the ratio between the two values is maximum. Judge that In determining the frame position, moving average processing is performed if the curve does not become a gentle curve due to the original sound spectrum.

  In this example, a watermarked voice signal is acquired (ST11), and after A / D conversion (ST12), FFT conversion is performed while synchronizing frames (ST13 to ST15), and a watermark signal is acquired.

  Note that the above two band spectrum grouping performed for embedding the additional signal is performed in the adjacent odd-order and even-order, but may be performed every other group. Furthermore, the transform from the time domain to the frequency domain can be performed using a DFT, DST, DCT, MDCT, etc., and can be performed by a transform including at least one of these transforms.

  In addition, as shown in FIG. 14C, in the case of FFT, the time domain frames are not overlapped and are converted from the time domain to the frequency domain (M = 0). In this case, if the spectrum intensity is changed and returned from the frequency domain to the time domain, a waveform change occurs due to the embedding of the signal in the frequency domain. May occur.

  In order to deal with such a situation, in MDCT, the frames are gently overlapped. That is, as shown in FIG. 14A, the frame is completely overlapped in the two regions in the time region. That is, by overlapping with the previous frame and applying a window function twice the frame update length N, M and N are equal when the frame length is N and the overlap width is M ( M = N). In this way, discontinuity is eliminated and sound quality is improved. Note that the overlap width may overlap in a part of the frame (M = N / 2) as shown in FIG. Such processing is not limited to MDCT, and can be similarly performed by any method such as FFT, DFT, DST, and DCT.

  In addition, the audio quality is degraded by dividing the audible band into multiple bands and embedding data in each band, or by adding processing such as not embedding in bands that have a large influence on sound, especially in the direct current region or the ultra low frequency region. Can be prevented. In broadcasting, the vertical synchronization signal of 60 Hz, the horizontal synchronization signal of 15.75 kHz jumps, and since there is a stereo pilot signal, etc. at a high frequency of about 19 kHz, this part is not embedded and is not taken into account when retrieving data. When processing is performed, better results are obtained in both sound quality and error rate.

  Prior to the embedding of information, it is desirable to perform a pre-processing that aligns the sum of the amplitudes of the spectra of the two groups, ensuring data reading. In particular, when a band in which no signal is embedded is provided, it is desirable to perform preprocessing to align the amplitude in the band or to exclude the band that is not embedded at the time of data extraction and obtain the total amplitude.

  In the present invention, by superimposing data on sound and extracting the data, it is possible to determine what is being broadcast on a television, radio, etc., what is recorded on a CD or DVD, More detailed information can be obtained from, for example, another computer connected to the Internet or the like, or the data itself can be presented. By superimposing text information such as subtitles and other audio data on the sound signal, the accessibility of the hearing impaired is improved, various information is presented at the time of disaster, etc., explanation using sub-speech and information in other languages Can also be inserted.

  In addition, it is possible to apply to the security field, such as authentication by sound, by changing the data to be superimposed even for sounds that can be heard equally.

It is a block diagram which shows the additional information embedding apparatus to the audio | voice data which concerns on embodiment of this invention. It is a flowchart which shows the additional information embedding method to the audio | voice data which concerns on embodiment of this invention. It is a block diagram which shows the additional information extraction apparatus to the audio | voice data which concerns on embodiment of this invention. It is a flowchart which shows the additional information extraction method to audio | voice data which concerns on embodiment of this invention. It is a graph of the audio | voice data time series explaining the outline | summary of the additional information embedding method to the audio | voice data of this invention. It is a figure which shows the band spectrum of the frequency series in the additional information embedding method to the audio | voice data which concerns on this invention. It is a figure which shows the FFT process of the additional information embedding method to the audio | voice data which concerns on embodiment. It is a figure which shows the pre-process of the additional information embedding method to the audio | voice data which concerns on embodiment. It is a figure explaining the embedding of the additional signal of the additional information embedding method to the audio | voice data which concerns on embodiment. It is a figure explaining the inverse FFT conversion of the additional information embedding method to the audio | voice data which concerns on embodiment. It is a figure explaining extraction of the additional signal of the additional information extraction method to the audio data according to the embodiment. It is a graph which shows the audio | voice data before adding the additional signal in the additional information embedding method to the audio | voice data which concerns on embodiment. It is a graph which shows the audio | voice data which added the additional signal in the additional information embedding method to the audio | voice data which concerns on embodiment. It is a graph which shows the state of the overlap in the additional information embedding method to the audio | voice data which concerns on embodiment.

Explanation of symbols

11 audio signal storage means 12 A / D conversion means 13 FFT conversion device 14 signal averaging means 15 additional signal embedding means 16 additional signal storage means 17 inverse FFT conversion means 18 DA converter 21 audio signal storage means 22 D / A conversion means 23 Frame synchronization means 24 FFT converter 25 Additional signal extraction means

Claims (22)

A method for embedding additional information in audio data, wherein the additional information signal is superimposed on a frequency amplitude characteristic of a spectral distribution of the audio data. Convert the discrete amplitude signal in the time domain of digitized audio data to the frequency domain,
Superimposing additional information on the frequency sequence data on the converted signal,
2. The method of embedding additional information in audio data according to claim 1, wherein the frequency series data on which the additional data is superimposed is converted into audio data. Extract the spectral distribution of audio data,
Dividing a predetermined frequency distribution region of the spectrum distribution into a plurality of bands;
The divided bands are combined into a plurality of data groups,
The gain of the plurality of data groups is increased or decreased based on the additional information value,
Organize multiple data groups after embedding re-addition information into a continuous spectral distribution,
3. The method for embedding additional information in audio data according to claim 1, wherein audio data to which additional information is added is created based on the spectrum distribution. 4. The method of embedding additional information in audio data according to claim 3, wherein the number of groups of data groups to be collected is 2, and 1-bit information is assigned according to the total amplitude of the data groups. 4. The method of embedding additional information in audio data according to claim 3, wherein preprocessing is performed to make the average value of the total amplitudes of the two data groups substantially equal. Conversion from time series data to frequency series data includes FFT (fast Fourier transform), DFT (discrete Fourier transform), DST (discrete sine transform), DCT (discrete cosine transform), MDCT (modified discrete cosine transform). 5. The method of embedding additional information in audio data according to claim 2, wherein the conversion is performed by conversion including at least one. 7. Audio data according to claim 1, wherein a plurality of areas of audio data are selected, processing is performed for each selected area, and additional information having a plurality of bits corresponding to the number of selected areas is embedded. Additional information embedding method. The method of embedding additional information in audio data according to any one of claims 1 to 7, wherein the frequency region in which the additional information is embedded is performed so as to avoid a band in which the reproduced sound has a great influence on the voice recognition of the listener. 9. The method for embedding additional information in audio data according to any one of claims 1 to 8, wherein the adjacent frames in the time domain are gently superimposed and converted when converting to the frequency series data. Extract the spectral distribution of audio data,
Dividing a predetermined frequency distribution region of the spectrum distribution into a plurality of bands;
The divided bands are combined into a plurality of data groups,
The magnitude of the plurality of data groups is determined for each data group,
A method for extracting additional information from audio data, wherein the additional information is extracted based on a magnitude relationship between the data groups. 11. The method for reading additional information from audio data according to claim 10, wherein the division and summarization of the audio frequency is performed in synchronization with an additional signal. An apparatus for embedding additional information in audio data, comprising: amplitude characteristic extracting means for extracting frequency amplitude characteristics of the spectral distribution of the audio data; and superimposing means for superimposing additional information signals on the frequency amplitude characteristics. The additional information superimposing means converts the discrete amplitude signal in the time domain of the digitized audio data into the frequency domain, superimposes the additional information on the frequency sequence data on the converted signal, and the frequency sequence on which the additional data is superimposed. 13. The apparatus for embedding additional information in audio data according to claim 12, further comprising conversion means for converting the data into audio data. Spectral distribution extraction means for extracting the spectral distribution of the audio data;
Band dividing means for dividing a predetermined frequency distribution area of the spectrum distribution into a plurality of bands;
Band dividing means for grouping the divided bands into a plurality of data groups;
Additional information adding means for increasing or decreasing the gain of the plurality of data groups based on the additional information value, and organizing the plurality of data groups after re-addition information embedding into a continuous spectrum distribution;
14. The apparatus for embedding additional information in audio data according to claim 12 or 13, further comprising audio data generating means for generating audio data to which additional information is added based on the spectrum distribution. The band dividing means sets the number of data groups to be grouped to 2,
15. The apparatus for embedding additional information in audio data according to claim 12, wherein the additional information adding means allocates 1-bit information according to the total amplitude of the data group. 16. The apparatus for embedding additional information in audio data according to claim 12, further comprising preprocessing means for making the average value of the total amplitude of the two data groups substantially equal. The amplitude characteristic extraction means converts time series data to frequency series data by FFT (fast Fourier transform), DFT (discrete Fourier transform), DST (discrete sine transform), DCT (discrete cosine transform), MDCT (deformation). 17. The apparatus for embedding additional information in audio data according to claim 12, wherein the conversion is performed by conversion including at least one of discrete cosine conversion. 18. The audio according to claim 12, wherein a plurality of frequency regions of audio data are selected, processing is performed for each selected region, and additional information having a plurality of bits corresponding to the number of selected regions is embedded. Device for embedding additional information in data. The apparatus for embedding additional information in audio data according to any one of claims 12 to 18, wherein a frequency region in which the additional information is embedded is performed in a range where the reproduced sound has a large influence on the voice recognition of the listener. 20. The additional information embedding in audio data according to any one of claims 12 to 19, wherein the amplitude characteristic extracting means performs conversion by superimposing frames adjacent in the time domain gently upon conversion to the frequency series data. Method.
Spectral distribution extraction means for extracting the spectral distribution of the audio data;
Band dividing means for dividing a predetermined frequency distribution area of the spectrum distribution into a plurality of bands;
And additional information extracting means for collecting the divided bands into a plurality of data groups, determining the magnitude of the plurality of data groups for each data group, and extracting additional information based on the magnitude relationship of the data groups. A device for extracting additional information from audio data. The apparatus for reading additional information from audio data according to claim 21, wherein the band dividing means comprises frame synchronization means for tuning to the audio frequency components and the combined additional signal.