JP2000242286A - Information embedding method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make to prevention of auditory change compatible with the resistance to sound processing by obtaining the change processing position of the value of contents which can not be sensed by a man or does not disturb contents reference by man, performing a processing such as deletion, addition, or no alteration at the change processing position, and embedding information. SOLUTION: A processing candidate point detection 4 is performed first to extract a flat area of source sound data 2 or a flat point right before abrupt variation and thus a processing candidate point is detected to generate processing candidate point data 5. Through watermark information embedding 6, watermark information data 3 are embedded in the sound source data 2 by deleting or adding data and watermark sound data 7 are outputted. A user processes the watermark sound data 7 having the watermark information embedded additionally (by copying, analog conversion, MPEG3 compression, cutting, etc.), to alter the data, thereby outputting inspected sound data 13.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital watermarking method for embedding various information such as a copyright, a sale date, and a sales destination in sound data.

[0002]

2. Description of the Related Art The following items are required for a digital watermark function for changing the data value of sound data and embedding information such as copyright depending on the manner of the change.

(1) A change in sound data cannot be recognized by a human. Alternatively, it can be determined that there is no problem in use.

(2) MPE is applied to sound data in which information is embedded.
Even if a process involving change of sound data such as G3 compression or analog conversion is performed, the embedded information can be detected without losing as much as possible.

In the conventional information embedding technology, as described in Nikkei Electronics (1997), No. 684, pp. 153 to 168, in order to satisfy these requirements, the type of value to be changed is devised. I was That is, a change is made to a value that is hardly noticeable and hardly disappears. For example, sound data is expressed in frequency, and a change is made to the coefficient of the middle frequency component.

[0006]

In the prior art, information has been embedded using the frequency characteristics of audio data. Therefore, if priority is given to preventing auditory changes,
It was necessary to make the value change small, and the tolerance of sound processing was small. On the other hand, when priority is given to the tolerance to sound processing, it is necessary to increase the value change, which causes an auditory change.
That is, it has been difficult to achieve both the prevention of auditory changes and the resistance to sound processing.

As described above, an object of the present invention is to provide a method for embedding information in sound data, which makes it possible to prevent both auditory changes and tolerate sound processing.

[0008]

Means for Solving the Problems As characteristics of sound data,
It has the following properties.

(1) Even if data is deleted or added in a flat area, almost no human can recognize it.

(2) In a flat area immediately before a rapid change, even if data is deleted or added, almost no human can recognize it.

It has been conceived to solve the above-mentioned problem by utilizing the above properties. That is, the above-mentioned problem is to detect the above-mentioned areas (1) and (2) and add or delete data in an information processing system having means for inputting sound data and means for embedding the sound data. Can be solved by embedding information.

[0012]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention are broadly classified into a part for embedding information in sound data and a part for extracting information from sound data in which information is embedded. The feature of the portion of the present invention in which the information is embedded in the audio data is to detect a flat region of the sound data or a flat point immediately before a sudden change, and delete or add the data to embed the information.

In the following, section (1) briefly describes the first embodiment of the present invention, and section (2) describes the case where original data is present at the time of reading and extracting information as the second embodiment. In the section 3), a case where there is no original data at the time of reading and extracting information will be described as a third embodiment.

(1) First Embodiment Hereinafter, referring to the functional configuration diagrams of FIGS. 1, 2, 3, and 4,
An embodiment of the present invention will be described. In this embodiment, sound data is targeted, and information is embedded by deleting or adding data.

FIG. 1 shows a watermark information embedding processing device 1.
2 shows the user addition processing device 11, FIG. 3 shows the watermark information detection processing device 21 when there is original sound data, and FIG. 4 shows the watermark information detection processing device 31 when there is no original sound data.

First, the functional configuration of FIGS. 1, 2, 3 and 4 will be described. The original sound data 2, the watermark information data 3, the watermark sound data 7, the inspected sound data 13, and the detected watermark information data 25 are realized by an external storage device. Processing candidate point detection 4, watermark information embedding 6, user addition processing 12, start position detection 22 (with original sound data), watermark information detection 24 (with original sound data), start position detection 32 (without original sound data), watermark information detection 34 (No original sound data)
Is realized by the CPU of the computer. Processing candidate point data 5,
The start position data 23 (with original sound data) and the start position data 33 (without original sound data) are data, and are realized by a storage device of a computer.

Next, the watermark information embedding process will be described with reference to FIG. First, a flat area of the original sound data 2 or a flat point immediately before a rapid change is extracted by the processing candidate point detection 4, and processing candidate points are detected to generate processing candidate point data 5. Then, the watermark information embedding 6 is used to delete or add data using the processing candidate point data 5 to embed the watermark information data 3 in the original sound data 2 and output the watermark sound data 7. Details will be described in Chapters (2) and (3).

Next, the user adding process will be described with reference to FIG. In this method, a user performs user addition processing 12 (copy, analog conversion, MPEG3 compression processing, cutout, etc.) on watermark sound data 7 in which watermark information is embedded, falsifies the data, and outputs inspection sound data 13.

Next, a description will be given of the watermark information detection processing in the case where there is the original sound data 2 with reference to FIG. First, processing candidate point data 5 of the original sound data 2 is obtained by processing candidate point detection 4.
Create Next, start position detection 23 detects start position data 23 using original sound data 2, sound data 13 to be inspected, and processing candidate point data 5. Next, the original sound data 2 and the inspected sound data 13 are detected by the watermark information detection 24.
Watermark information data 25 using the
Output. Details will be described in Chapter (2).

Next, a description will be given of the watermark information detection processing in the case where there is no original sound data 2 with reference to FIG. First, start position data 33 is detected by the start position detection 32 using the test sound data 13. Next, the watermark information detection 34 outputs the detected watermark information data 25 using the start position data 33 and the test sound data 13. Details will be described in (3).

(2) Second Embodiment (When Original Data Exists at Reading and Retrieving Information) A second embodiment of the present invention will be described below in two sections. Section (2-1) describes the watermark embedding processing, and Section (2-2) describes the watermark information detection processing.

(2-1) Watermark Embedding Processing An embodiment of the watermark embedding processing will be described with reference to FIG.
FIG. 5 is a flowchart of the watermark embedding process. Hereinafter, the procedure will be described using the numbers of the flowchart of FIG. 5 as step numbers.

Step 41: Calculate the absolute value of the difference between the original sound data and create the difference data.

Step 42: Processing candidate points are determined based on the difference data, and processing candidate point data 5 is created. Details are shown in section (2-1-1).

Step 43: The watermark information is embedded in the original sound data on the basis of the watermark information at the processing candidate points of the original sound data to create watermark sound data. See (2-1-
See section 2).

(2-1-1) Method for Creating Processing Candidate Point Data A method for creating processing candidate point data will be described with reference to FIGS. FIG. 6 is a flow chart of the processing for creating the processing candidate point data. Hereinafter, the procedure will be described with the numbers in the flow chart of FIG. 6 as step numbers.

However, it is assumed that the following parameters are set in advance.

Flat point detection value 62: Sound data value for detecting flat point 66 Rapid change point detection value 63: Sound data value for detecting sudden change point 67 Flat continuous value 64: Detecting flat area 68 Continuous value Processing candidate point interval 65: Interval value of processing candidate point 72 Step 51: A point smaller than flat point detection value 62 is detected from difference data 61 as flat point 66, and a point larger than sudden change point detection value 63 is detected as sudden change point 67. . When the substantially flat point 66 is more continuous than the flat continuous value 64, the flat area 68 is extracted.

Step 52: The flat point 70 at the center of the substantially flat area 68 and the flat point 71 immediately before the rapidly changing area 69 are detected as processing candidate points.

Step 53: rough machining candidate target points 70 and 7
A point having an interval of 65 or more from 1 is determined as a candidate processing point 72, and processing candidate point data 5 is created.

(2-1-2) Watermark Information Embedding Processing The watermark information embedding processing method will be described with reference to FIGS. FIG. 8 is a flowchart of the watermark information embedding process. Hereinafter, the procedure will be described by using the numbers in the flowchart of FIG. 8 as step numbers.

Step 81: The Flag and the transfer position are initialized.

Step 82: A candidate processing point position 104 is extracted from the approximate candidate processing point data 5.

Step 83: The original sound data 102 from the transfer position to before the processing candidate point position 104 is transferred to the watermark sound data 103.

Step 84: One bit is extracted from the watermark information data 101. When it becomes the last bit of the watermark information data 101, it returns to the beginning and extracts it.

Step 85: If the information extracted in step 84 is 0, the process jumps to step 88; if not, the process jumps to step 87.

Step 86: If Flag is 0, jump to step 89; if not, jump to step 87. .

Step 87: Flag is set to 0, original sound data 102 at processing candidate point position 104 and processing candidate point position 1
After transferring the average value 107 of the original sound data 102 which is one prior to 04 to the data addition position 106 of the watermark sound data 103, the data 105 of the original sound data 102 at the processing candidate point position 104 is transferred to the watermark sound data 103. This processing is data addition processing.

Step 88: The data 105 of the original sound data 102 at the processing candidate point position 104 is transferred. This is without any processing.

Step 89: Flag is set to 1. This process is a data reduction process.

Step 90: The next position of the processing candidate point position 104 is set as the transfer position.

Step 91: If the processing candidate point position 104 remains in the rough processing candidate point data 5, the process returns to step 82.

Step 92: Original sound data 1 from transfer position
The data of the original sound data 102 up to the end point 02 is transferred to the watermark sound data 103.

(2-2) Watermark Information Detection Process An embodiment of the watermark information detection process will be described with reference to FIG. FIG. 10 is a flowchart of the watermark information detection process. Hereinafter, the procedure will be described using the numbers in the flowchart of FIG. 10 as step numbers.

Step 111: Create processing candidate point data from the original sound data. The creation method is the same as (2-1-1).

Step 112: From the original sound data and the test sound data, the starting positions of the original sound data and the test sound data are detected. This processing is necessary when the test sound data is created by the cut processing. If the cut processing is not performed, the start positions of the original sound data and the test sound data are the start positions. Details are shown in section (2-2-1).

Step 113: The detected watermark information data is detected from the start positions of the original sound data and the sound data to be inspected. Details are given in section (2-2-3).

(2-2-1) Start Position Detection Method A start position detection method will be described with reference to FIGS. FIG. 11 is a flowchart of the start position detecting process. Hereinafter, the procedure will be described by using the numbers in the flowchart of FIG. 11 as step numbers.

However, it is assumed that the following parameters are set in advance.

Number of processing candidate point skips: Number of processing candidate point skips Detection search range value 139: Search range of start position Step 121: Original sound data 131 and test sound data 1
32, the approximate start position 133 is detected. In the detection, a position where the head of the test sound data 132 matches the original sound data 131 is detected by comparing sounds.

Step 122: Extract original sound search data 135 obtained by skipping the processing candidate point 134 by the processing candidate point skip number from the approximate start position 133 of the original sound data 131.

Step 123: The similarity is calculated using the original sound search data 135 in the range of the detection search range value 139 from the head of the sound data 132 to be inspected, a position having a large similarity is detected, and the original sound start position 137 and the original sound start position 137 are detected. The inspection sound start position 138 is detected. Details of the method of calculating the similarity will be described in the section shown in section (2-2-2).

(2-2-2) Method of Calculating Similarity The calculation of similarity is performed using any one of the similarity calculation methods described below.

Correlation Value Method A correlation value between the original sound data 131 and the test sound data 132 is calculated, and a larger value is regarded as having a higher similarity.

Sum of squares method of difference The sum of squares of the difference between the original sound data 131 and the test sound data 132 is calculated, and a small value is regarded as a large similarity.

Absolute Value Method of Difference The sum of the absolute value of the difference between the original sound data 131 and the test sound data 132 is calculated, and a smaller value is regarded as having a higher similarity.

In the matching method, the value obtained by subtracting the average value of the original sound data 131 from the original sound data 131 instead of the original sound data 131 and the sound to be inspected from the sound data to be inspected 132 instead of the sound data to be inspected 132 A value obtained by subtracting the average value of the data 132 is used.

(2-2-3) Watermark Information Data Detection Method An embodiment of the watermark information data detection processing will be described with reference to FIGS. FIG. 13 is a flowchart of the watermark information data detection process. The procedure will be described below using the numbers in the flowchart of FIG. 13 as step numbers.

Step 141: The data length from the start position 169 of the original sound data 161 to the next processing candidate point 166 is detected.

Step 142: The start position 164 of the original sound data 161 is set at the original sound detection position, and the start position 165 of the test sound data 162 is set at the test sound position.

Step 143: The original sound detection position and the inspected sound position are shifted by the data length +1 detected in step 142.

Step 144: The data length from the original sound detection position to the next processing candidate point 166 of the original sound data 161 is detected.

Step 145: Extract original sound data 161 having a data length of -1 detected in step 144 from the original sound detection position as original sound check data 167.

Step 146: Calculate the similarity between the test sound data 162 and the original sound check data 167 from the test sound position, the test sound position -1 and the test sound position +1. Note that the calculation of the similarity uses (2-2-2).

Step 147: When the similarity calculated in step 146 from the position of the sound to be inspected is the largest Step 1
Jump to step 49, otherwise jump to step 148.

Step 148: If the similarity calculated in step 146 from the test sound position -1 is the largest, the process jumps to step 150; otherwise, the process jumps to step 151.

Step 149: In the watermark information embedding processing, 0 is output to the detected watermark information data 163 because the processing of the processing candidate point is not changed. Further, the original sound detection position changes by the data length detected in step 144. Also, the position of the sound to be inspected changes by the data length detected in step 144.

Step 150: 1 is output to the detected watermark information data 163 because the processing at the processing candidate point in the watermark information embedding processing is the data deletion point. Further, the original sound detection position changes by the data length detected in step 144. In addition, the position of the sound to be inspected changes by the data length detected in step 144-1.

Step 151: Since the processing of the processing candidate point in the watermark information embedding processing is a data additional data point (additional data 168 is added), one is added to the detected watermark information data 163.
Is output. Further, the original sound detection position changes by the data length detected in step 144. Further, the position of the sound to be inspected is shifted by the data length detected in step 144 + 1.

Step 152: If the position of the sound to be inspected does not exceed the sound data 162 to be inspected, step 144
Fly to.

Step 152: detected watermark information data 1
63 is output.

(3) Third Embodiment (In the case where there is no original data at the time of reading and extracting information) As shown in FIG. 15, the central value 172 (127 in the case of 1-byte data) is set to 0 as a feature of the sound data 171. In this case, when the sum of the cut-out areas is calculated to be constant, the sum is close to zero. Using this feature, the original sound data is divided into divided data size 173 (for example, 440 data when sound data of 44 kHz is divided in 0.01 seconds), and data is deleted such that the sum between the divided data becomes + or-. By adding data or data, watermark information can be embedded. The detection of the watermark information can be realized by detecting the division position of the divided data and determining whether the sum of the divided data thereafter is + or-.

Hereinafter, a third embodiment of the present invention will be described.
The explanation is divided into sections. Section (3-1) describes the watermark embedding processing, and section (3-2) describes the watermark information detection processing.

However, it is assumed that the following parameters are set in advance.

Partition Data Size: Partition Data Size (3-1) Watermark Embedding Processing An embodiment of the watermark embedding processing will be described with reference to FIG. FIG. 16 is a flowchart of the watermark embedding process. Hereinafter, the procedure will be described using the numbers in the flowchart of FIG. 16 as step numbers.

Step 181: Calculate the absolute value of the difference between the original sound data and create the difference data.

Step 182: The original sound data and the difference data are divided into divided data sizes. Hereinafter, the data divided into the division data sizes is referred to as division data.

Step 183: Based on the difference data, a processing candidate point in the section data is determined, and processing candidate point data is created. However, the number of processing candidate points with data "+" and the number of processing candidate points with data "-" are the same. For details, see (3-1-
See section 1).

Step 184: 1 from watermark information data
Take out the bits. When the watermark information data becomes the last bit, return to the beginning and extract it.

Step 185: If the information extracted in step 184 is 0, the process jumps to step 187; otherwise, the process jumps to step 186.

Step 186: If the data of the processing candidate point is +, data addition processing is performed, and if the data of the processing candidate point is-, data deletion processing is performed so that the sum of the data in the section data becomes +. Details are shown in section (3-1-2).

Step 187: If the data of the processing candidate point is-, the data addition processing is performed, and if the data of the processing candidate point is +, the data deletion processing is performed so that the sum of the data in the section data becomes-. Details are shown in section (3-1-2).

Step 188: The section data processed in step 186 or 187 is output as watermark sound data.

Step 189: If the section data remains, the process jumps to step 183.

Step 190: The remaining original sound data is output as watermark sound data.

(3-1-1) Method for Creating Processing Candidate Point Data A method for creating processing candidate point data will be described with reference to FIGS. FIG. 17 is a flowchart of the processing for creating candidate processing point data. The procedure will be described below using the numbers in the flowchart of FIG. 17 as step numbers.

However, it is assumed that the following parameters are set in advance.

Flat point detection value 212: Sound data value for detecting flat point 215 Rapid change point detection value 213: Sound data value for detecting rapid change point 216 Processing candidate point interval 214: For detecting processing candidate point Step 201: Flat point detection value 2 from difference data 211
A point smaller than 12 is a flat point 215 and a sudden change point detection value 213.
A larger point is detected as a sudden change point 216. When the substantially flat point 215 is continuous from the processing candidate point interval 214, it is extracted as a flat area 217.

Step 202: The roughly flat area 217 is divided at the processing candidate point interval 214, and the flat point 219 in the center and the flat point 220 immediately before the almost flat sudden change area 218 are detected as processing candidate target points.

Step 203: Extract the processing candidate target points with + data and the processing candidate target points with-as small as the number of processing candidate target points with + data and the number of processing candidate target points with-data. To create machining candidate point data.

(3-1-2) Watermark Information Embedding Processing The watermark information embedding method will be described with reference to FIGS. 20, 21 and 22. FIG. 20 shows original sound data 2 in the section data.
The processing candidate point 232 on 31 is shown. FIG. 21 shows that, if the data of the processing candidate point 232 is +, the data of the processing candidate point 232 is added so that the sum in the division data becomes +, and if the data of the processing candidate point 232 is-, the processing is performed. The data of the candidate point 232 is deleted to show the section data. FIG.
Is the processing candidate point 2 if the data of the processing candidate point 232 is-so that the sum in the section data becomes-
32 data is added, and the data of the machining candidate point 232 is +
If so, the data of the processing candidate point 232 is deleted to indicate the section data. Thus, if the watermark information data is 0, FIG.
If the watermark information data is 1, as shown in FIG. 21, the processing in the divided data is performed as shown in FIG. 21 to embed the watermark information.

(3-2) Watermark Information Detection Processing An embodiment of the watermark embedding processing will be described with reference to FIG. FIG. 23 is a flowchart of the watermark embedding process. Hereinafter, the procedure will be described by using the numbers in the flowchart of FIG. 23 as step numbers.

Step 241: The test sound data is evaluated from the position shifted by 1/16 of the section data size, and the start position is detected. Details are given in section (3-2-1). However, when the test sound data has not been cut out, the start position is the head position of the test sound data.

Step 242: The test sound data is divided from the start position by the division data size, and the detection watermark information is extracted based on whether the sum of the data in the division data is plus or minus to detect the information data. Details are shown in section (2-2-2).

(3-2-1) Start Position Detection Method A start position detection method will be described with reference to FIG. First, the watermark information is embedded so that the sum of the data values in the segmented data becomes + or-. If the amount of segmented data is sufficiently large, the following assumption is made. here,
251 is a start position at the time of setting watermark information, 252 is a start position at the time of deviation from the information setting, 253 is a section data size, and 254 is an average value of data values in the section data. As shown in FIG. 24, the sum of the data values in the segmented data is small at a place where the sign is reversed in the +/− direction, and is smaller when the deviation of the start position is large. Therefore, by accumulating the absolute value of the sum in the divided data by shifting the start position, the start position having the largest cumulative value becomes the start position closest to the start position at the time of setting the information.

Next, the processing procedure will be described with reference to the flowchart of the start position detecting method processing of FIG. 25, where the numbers in the flowchart of FIG. 25 are step numbers.

Step 261: Set partition data size to 16
Divide equally and set the start position.

Step 262: The to-be-tested sound data is divided into divided data sizes from the start position set in step 261.

Step 263: The absolute value of the sum of the data values in the section data is accumulated.

Step 264: If all set start positions have not been completed, the process jumps to step 262.

Step 265: The start position having the maximum total value calculated in step 263 is output as start position data.

(3-2-2) Watermark Information Data Detection Method An embodiment of the watermark information data detection processing will be described with reference to FIGS. 26 and 18. FIG. 26 is a flowchart of the watermark information data detection process. The procedure will be described below using the numbers in the flowchart of FIG. 26 as step numbers.

Step 271: From the start position 284 of the start position data, the to-be-inspected sound data is divided into divided data sizes.

Step 272: If the sum (average value 282) of the data values in the section data 281 is minus, then step 27
Fly to 4.

Step 273: Detection watermark information data 2
0 is output to 83.

Step 274: Detection watermark information data 2
Output 1 to 83.

Step 275: If the division data is not completed, the flow jumps to step 272.

Step 276: Detected watermark information data 2
83 is output.

[0109]

According to the present invention, in embedding information into sound data, it is possible to prevent deterioration and to attain both the durability of embedded information.

[Brief description of the drawings]

FIG. 1 is a functional configuration diagram of a watermark information embedding processing device according to a first embodiment of the present invention.

FIG. 2 is a functional configuration diagram of a user's account addition processing device according to the first embodiment of the present invention.

FIG. 3 is a functional configuration diagram of a watermark information detection processing device with original sound data according to the first embodiment of the present invention.

FIG. 4 is a functional configuration diagram of a watermark information detection processing device without original sound data according to the first embodiment of the present invention.

FIG. 5 is a flowchart of watermark information embedding processing according to a second embodiment of the present invention.

FIG. 6 is a flowchart of processing candidate point detection processing.

FIG. 7 is a specific example of processing candidate point detection processing.

FIG. 8 is a flowchart of how to create watermark information.

FIG. 9 shows a specific example of watermark information embedding processing.

FIG. 10 is a flowchart of a watermark information detection process according to a second embodiment of the present invention.

FIG. 11 is a flowchart of a start position detection process.

FIG. 12 is a specific example of a start position detection process.

FIG. 13 is a flowchart of a method for detecting watermark information.

FIG. 14 is a specific example of a method of detecting watermark information.

FIG. 15 is an explanatory diagram of sound data characteristics according to a third embodiment of the present invention.

FIG. 16 is a flowchart of watermark information embedding processing according to a third embodiment of the present invention.

FIG. 17 is a flowchart of processing candidate point detection processing.

FIG. 18 is a specific example of a method of detecting watermark information.

FIG. 19 is a specific example of processing candidate point detection processing.

FIG. 20 shows a specific example of watermark information embedding processing.

FIG. 21 shows a specific example of watermark information embedding processing.

FIG. 22 shows a specific example of watermark information embedding processing.

FIG. 23 is a flowchart of watermark information detection processing according to a third embodiment of the present invention.

FIG. 24 is a specific example of a start position detection process.

FIG. 25 is a specific example of a start position detection process.

FIG. 26 is a flowchart of how to detect watermark information.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... watermark information embedding apparatus, 2 ... original sound data, 3 ... watermark information data, 4 ... processing candidate point detection processing, 5 ... processing candidate point data, 6 ... watermark information embedding processing, 7 ... watermark sound data.

Claims (1)

[Claims] In an information processing system having means for inputting content and means for embedding information in the content, a change processing position of a value of the content which cannot be perceived by a human or hinders a human from referring to the content is obtained. An information embedding method characterized in that information is embedded by performing processing of deletion, addition, or no change at the changed processing position.