CN113470666A - Reversible robust medical audio method based on two-stage embedding - Google Patents

Reversible robust medical audio method based on two-stage embedding Download PDF

Info

Publication number
CN113470666A
CN113470666A CN202110687228.2A CN202110687228A CN113470666A CN 113470666 A CN113470666 A CN 113470666A CN 202110687228 A CN202110687228 A CN 202110687228A CN 113470666 A CN113470666 A CN 113470666A
Authority
CN
China
Prior art keywords
audio
domain
watermark
frequency
embedding
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110687228.2A
Other languages
Chinese (zh)
Other versions
CN113470666B (en
Inventor
张小瑞
孙逊
孙星明
宋爱国
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Information Science and Technology
Original Assignee
Nanjing University of Information Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing University of Information Science and Technology filed Critical Nanjing University of Information Science and Technology
Priority to CN202110687228.2A priority Critical patent/CN113470666B/en
Publication of CN113470666A publication Critical patent/CN113470666A/en
Application granted granted Critical
Publication of CN113470666B publication Critical patent/CN113470666B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L19/018—Audio watermarking, i.e. embedding inaudible data in the audio signal

Landscapes

  • Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Signal Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
  • Signal Processing For Digital Recording And Reproducing (AREA)

Abstract

The invention discloses a reversible robust medical audio method based on two-stage embedding, which comprises the following steps: (1) the original audio is converted into two independent embedded domains by a frequency domain transform function: a low frequency embedded domain and a high frequency embedded domain; (2) embedding a watermark in a low-frequency embedded domain by adopting a robust watermark algorithm to form a new low-frequency embedded domain; (3) embedding the watermark error between the new low-frequency embedded domain and the low-frequency embedded domain into the high-frequency embedded domain to form a new high-frequency embedded domain; (4) generating the watermark-containing audio by inverse transformation of the frequency domain transformation function for the new low-frequency embedded domain and the new high-frequency embedded domain; (5) and extracting watermark information from the audio containing the watermark. The invention can effectively extract the watermark information and restore the complete audio frequency under the condition that the audio frequency is not attacked, and is used for the integrity and authenticity authentication of the medical data; the robustness of the watermark is enhanced by using the continuity of the audio and the correlation between sampling points, a small number of sampling points are modified in the reversible watermark, and the audio distortion is reduced.

Description

Reversible robust medical audio method based on two-stage embedding
Technical Field
The invention relates to the field of digital watermarking, in particular to a reversible robust medical audio method based on two-stage embedding.
Background
With the progress of internet and communication technology, remote medical treatment and non-contact medical treatment are widely regarded in the medical field, however, in the current complicated network era, medical data may be maliciously tampered in the transmission process of remote medical treatment, so that data distortion or illegal acquisition causes privacy disclosure of patients. Even small levels of distortion in medical data can affect a physician's diagnosis of a patient's condition, with serious consequences. When the traditional robust watermark and the fragile or semi-fragile watermark extract the watermark, the carrier data can be permanently distorted, so that a doctor cannot effectively diagnose.
Disclosure of Invention
The purpose of the invention is as follows: in view of the above problems, it is an object of the present invention to provide a reversible robust medical audio method based on two-stage embedding.
The technical scheme is as follows: the reversible robust medical audio method based on the two-stage embedding comprises the following steps:
(1) converting original audio into two independent embedding domains including a low frequency embedding domain and A by a frequency domain transformation function FlHigh frequency embedded domain Ah；
(2) Embedding a domain A in a low frequency by adopting a robust watermarking algorithmlA robust watermark w1 is embedded in the medium to form a new low-frequency embedded domain
Figure BDA0003125178960000011
(3) Will new low frequency audio
Figure BDA0003125178960000012
And A of the original audiolEmbedding the watermark error E as reversible watermark information w2 into the high-frequency embedding domain AhIn, forming a new high frequency embedded domain
Figure BDA0003125178960000013
(4) Embedding new low frequency into domain
Figure BDA0003125178960000014
And a new high frequency embedded domain
Figure BDA0003125178960000015
By frequency domain transformation functionsIs inverse transformation F-1Generating watermark-containing Audio Aw；
(5) For watermark-containing audio AwAnd extracting watermark information.
Further, the audio conversion process in step 1 is as follows:
the original audio A is represented as a set of sample points, denoted as { a }1，a2，a3,a4,...,a2g-1,a2gDecomposing original audio A into low-frequency embedded domain A by wavelet transformlAnd a high frequency embedded domain AhWherein, the sampling point pair is processed by a frequency domain transformation function F, and the expression is as follows:
Figure BDA0003125178960000016
where g represents half the number of sampling points, g being an integer, alRepresenting a low frequency embedded domain signal, ahRepresenting a high frequency embedded domain signal; low frequency embedded domain AlAnd a high frequency embedded domain AhThe lengths are the same.
Further, step 2 includes the following steps in the process of embedding the robust watermark:
(201) embedding low frequencies into domain AlUniformly dividing the audio into sub-audios which are not overlapped and have the same length, wherein each sub-audio has 2n sampling points, n represents half of the number of the sampling points of one sub-audio, n is an integer, and the sampling point of each sub-audio is marked as al(1),al(2),......,al(2n), in order to ensure information security, defining a random mapping relation:
Figure BDA0003125178960000021
the positions of the original sampling points are orderly disturbed,
Figure BDA0003125178960000022
a sequence of positions representing the sample points after mapping;
definition of
Figure BDA0003125178960000023
To representLow frequency embedded domain AlIn the sub-audio of
Figure BDA0003125178960000024
I is more than or equal to 1 and less than or equal to 2 n;
(202) embedding a 1-bit watermark w1 for each sub-audio of the low-frequency embedded domain, wherein w1 belongs to {0,1}, and the embedding formula is as follows:
Figure BDA0003125178960000025
wherein
Figure BDA0003125178960000026
Representing the low-frequency embedded domain A after embedding a watermarklIn the sub-audio of
Figure BDA0003125178960000027
A sampling point value of
Figure BDA0003125178960000028
The composed set is represented as a new low frequency embedded domain
Figure BDA0003125178960000029
μ is watermark embedding strength, dlRepresenting the low frequency embedded domain AlThe difference value of the sampling point sets of the front part and the rear part of the sub audio is expressed as follows:
Figure BDA00031251789600000210
after embedding the robust watermark, the new low frequency embedded domain
Figure BDA00031251789600000211
Difference value of sampling point sets of front and back parts of sub-audio
Figure BDA00031251789600000212
The modification is as follows:
Figure BDA00031251789600000213
(203) the sampling point value is changed to generate overflow, the sampling point mark sequence which overflows after being changed is restored to the original sampling point value, a new sampling point value is reserved for sampling points which do not overflow, and the extraction of the watermark is not influenced because the sampling point value of the overflow point has small change amount and the sampling points which overflow are generated little.
Further, step 3, the watermark error is denoted by E, and the factors for generating the watermark error include the watermark embedding strength mu and the low frequency embedding area AlThe difference value set D ═ D of sampling point sets of the front part and the back part of the sub-audiol(1),dl(2) ,.. }, wherein dl(1),dl(2) Respectively representing the low frequency embedded domain AlD of the first and second sub-audiolThe expression is:
Figure BDA00031251789600000214
wherein
Figure BDA00031251789600000215
Representing an additive binary operation.
Further, the step 3 embedding process includes:
(301) embedding high frequencies into domain AhUniformly dividing the audio into sub-audio frequencies which are not overlapped and have equal length, wherein sampling points in each sub-audio frequency are divided into M groups, and M represents the number of groups in one sub-audio frequency;
each set comprising 2 sampling points per group,
Figure BDA0003125178960000031
a qth sampling point representing the kth sampling point group in the sub-audio of the pth high frequency embedding domain, q ∈ {1,2 };
Figure BDA0003125178960000032
represents the pth high frequency embedded domain AhOf the k-th group of samples of the sub-audio, k being an integerThe expression is:
Figure BDA0003125178960000033
the difference statistic of the sub-audio of the pth high-frequency embedded domain is called S (p), and the expression:
Figure BDA0003125178960000034
(302) by changing
Figure BDA0003125178960000035
And (3) realizing the change of S (p), so that the watermark is embedded reversibly, wherein the embedding expression is as follows:
Figure BDA0003125178960000036
wherein B represents the amount of change in S (p),
Figure BDA0003125178960000037
means not exceeding
Figure BDA0003125178960000038
Is the largest integer of (a) to (b),
Figure BDA0003125178960000039
to represent
Figure BDA00031251789600000310
The changed value; s (p)' represents the changed value of s (p), and the expression is:
Figure BDA00031251789600000311
selecting secret key T > | SmaxWhere S represents a consonant of the high frequency embedded domainThe overall difference statistic for frequency, change B of s (p), is calculated as:
Figure BDA00031251789600000312
(303) reversible embedding of watermark sequences into high frequency embedding domain A by changing S (p) histogram of moving difference statisticshIn the method, 1 bit watermark information is embedded in the sub-audio of each high-frequency embedded domain, and w (p) represents AhThe watermark embedded in the sub audio of the pth high-frequency embedding domain, w (p) is E, and E represents a watermark error; when the watermark is 0, s (p) is unchanged, and when the watermark is 1, s (p)' ═ s (p) + B, i.e.
Figure BDA00031251789600000313
The difference statistic S ═ { S (p) |1 ≦ p ≦ N }, where N is the high-frequency embedding region AhThe number of neutron audios; order to
Figure BDA00031251789600000314
Wherein α (k) represents
Figure BDA0003125178960000041
By the q-th sampling point of the k-th sampling point group in the sub-audio of the p-th high-frequency embedded domain
Figure BDA0003125178960000042
Implementing integer transforms
Figure BDA0003125178960000043
To effect a change in s (p); the integer transform expression is:
Figure BDA0003125178960000044
wherein
Figure BDA0003125178960000045
Is that
Figure BDA0003125178960000046
The value after integer transformation is composed of
Figure BDA0003125178960000047
The constructed set represents the new high frequency embedded domain
Figure BDA0003125178960000048
(304) After transformation
Figure BDA0003125178960000049
The audio may not be in the range of the original sampling point value, and the audio needs to be preprocessed before the watermark is embedded, so as to prevent the overflow phenomenon from being generated, which is obtained by the following formula:
Figure BDA00031251789600000410
since k is more than or equal to 1 and less than or equal to M,
Figure BDA00031251789600000411
therefore, the method comprises the following steps:
Figure BDA00031251789600000412
order to
Figure BDA00031251789600000413
Wherein σ represents
Figure BDA00031251789600000414
The maximum value can be obtained by firstly passing through the sampling point with the embedded watermark and marking the sampling value which is not in the range of the original sampling point value,and then, the marked sampling point values are adjusted into the original sampling point values, the values lower than the lower limit of the original sampling point values are adjusted to the lower limit, the values higher than the upper limit of the original sampling point values are adjusted to the upper limit, the range of the sampling point values is large, sigma is small, overflowed sampling points are few, and the influence on the audio quality is small. The range of sample point values is large and σ is small, the overflow sample points are few, and the influence on the audio quality is small.
Further, the process of generating the watermark-containing audio in the step 4 comprises the following steps: embedding the new low frequency in step 2 into the domain
Figure BDA00031251789600000415
And the new high frequency embedded domain in step 3
Figure BDA00031251789600000416
Inverse transformation F by a transformation function-1Reconstructing watermark-containing Audio Aw。
Further, the step 5 of extracting comprises the following steps:
(501) watermark-containing audio A through frequency domain function FwDecomposition into two independent embedded domains, including a new low frequency embedded domain
Figure BDA0003125178960000051
And a new high frequency embedded domain
Figure BDA0003125178960000052
(502) Reversible watermark information w2 is extracted by shifting the differential histogram and stored,
Figure BDA0003125178960000053
is recovered to be Ah；
(503) From
Figure BDA0003125178960000054
The robust watermark w1 is extracted, and the reversible watermark information w2 is used for converting the robust watermark w1 into the reversible watermark information
Figure BDA0003125178960000055
Is recovered to be Al；
(504) Inverse transformation F by a transformation function-1A is to behAnd AlReverting to the original audio a.
Has the advantages that: compared with the prior art, the invention has the following remarkable advantages:
1. the invention is a reversible robust watermarking algorithm, can effectively extract watermarking information and restore complete audio under the condition that the audio is not attacked, and is used for the integrity and authenticity authentication of medical data;
2. the robustness of the watermark is enhanced by using the continuity of the audio and the correlation between sampling points, a small number of sampling points are modified in the reversible watermark, and the audio distortion is reduced;
3. the robust watermark and the reversible watermark are respectively embedded by utilizing two independent embedding domains, so that the influence of the reversible watermark on the robust watermark is effectively reduced, and the robustness is improved.
Drawings
Fig. 1 is a diagram of a watermark embedding framework of the present invention;
fig. 2 is a diagram of a watermark extraction framework of the present invention.
Detailed Description
The reversible robust medical audio method based on two-stage embedding described in this embodiment includes:
(1) converting the audio into two independent embedding domains including a low frequency embedding domain and a high frequency embedding domain by a frequency domain transformation function;
the original audio A is represented as a set of sample points, denoted as { a }1,a2,a3,a4,...,a2g-1,a2gDecomposing original audio A into low-frequency embedded domain A by wavelet transformlAnd a high frequency embedded domain AhWherein the sampling point pairs are processed by a transformation function F, and the expression is:
Figure BDA0003125178960000056
where g represents half the number of sampling points, g being an integer, alRepresenting a low frequency embedded domain signal, ahRepresenting high frequenciesAn embedded domain signal;
low frequency embedded domain AlAnd a high frequency embedded domain AhThe lengths are the same.
(2) And embedding the robust watermark in the low-frequency embedded domain by adopting a robust watermark algorithm to form a new low-frequency embedded domain. Fig. 1 is a diagram of a watermark embedding framework.
(201) Embedding low frequencies into domain A, as shown in Table 1lUniformly dividing the audio into sub-audios which are not overlapped and have the same length, wherein each sub-audio has 2n sampling points, n represents half of the number of sampling points of a character audio, n is an integer, and the sampling point of each sub-audio is marked as al(1),al(2)，......，al(2n), defining a random mapping relation:
Figure BDA0003125178960000057
the positions of the original sampling points are orderly disturbed,
Figure BDA0003125178960000058
a sequence of positions representing the sample points after mapping;
definition of
Figure BDA0003125178960000061
Representing the low frequency embedded domain AlIn the sub-audio of
Figure BDA0003125178960000062
I is more than or equal to 1 and less than or equal to 2 n;
TABLE 1 p sub-Audio Frames for the Low frequency Embedded Domain
Figure BDA0003125178960000063
(202) Embedding a 1-bit watermark w1, w 1E {0,1} for each sub-audio, the embedding formula is as follows:
Figure BDA0003125178960000064
wherein
Figure BDA0003125178960000065
Representing the low-frequency embedded domain A after embedding a watermarklIn the sub-audio of
Figure BDA0003125178960000066
A sampling point value of
Figure BDA0003125178960000067
The composed set is represented as a new low frequency embedded domain
Figure BDA0003125178960000068
μ is the embedding strength of the watermark, dlRepresenting the low frequency embedded domain AlThe difference value of the sampling point sets of the front part and the rear part of the sub audio is expressed as follows:
Figure BDA0003125178960000069
after embedding the robust watermark, the new low frequency embedded domain
Figure BDA00031251789600000610
The difference between the sampling points of the front and the rear parts of the sub-audio
Figure BDA00031251789600000611
The modification is as follows:
Figure BDA00031251789600000612
(203) the sampling point value is changed to generate overflow, the sampling point mark sequence which overflows after being changed is restored to the original sampling point value, a new sampling point value is reserved for sampling points which do not overflow, and the extraction of the watermark is not influenced because the sampling point value of the overflow point has small change amount and the sampling points which overflow are generated little.
(3) And embedding the watermark error between the new low-frequency audio and the original audio into a high-frequency embedded domain to form a new high-frequency embedded domain, wherein the watermark error serves as reversible watermark information.
The watermark error is denoted by E, and the factors for generating the watermark error include the watermark embedding strength mu and the low frequency embedding area AlThe difference value set D ═ D of sampling point sets of the front part and the back part of the sub-audiol(1),dl(2) ,.. }, wherein dl(1),dl(2) Respectively representing the low frequency embedded domain AlD of the first and second sub-audiolThe expression is:
Figure BDA00031251789600000613
wherein
Figure BDA00031251789600000614
Representing an additive binary operation.
The specific embedding process comprises the following steps:
(301) embedding high frequency into Domain A, as shown in Table 2hUniformly dividing the audio into sub-audios which are not overlapped and have equal length, wherein sampling points in each sub-audio are divided into M groups, and M represents the number of groups in the sub-audio of a high-frequency embedded domain;
each set comprising 2 sampling points per group,
Figure BDA00031251789600000615
a qth sampling point representing the kth sampling point group in the sub-audio of the pth high frequency embedding domain, q ∈ {1,2 };
Figure BDA0003125178960000071
representing the difference of the sampling points of the kth sampling point group of the sub audio of the pth high-frequency embedded domain, and the expression is as follows:
Figure BDA0003125178960000072
the difference statistic of the sub-audio of the pth high-frequency embedded domain is called S (p), and the expression:
Figure BDA0003125178960000073
TABLE 2 p sub-Audio Frames for high frequency Embedded Domain
Figure BDA0003125178960000074
(302) By changing
Figure BDA0003125178960000075
And (3) realizing the change of S (p), so that the watermark is embedded reversibly, wherein the embedding expression is as follows:
Figure BDA0003125178960000076
wherein B represents the amount of change in S (p),
Figure BDA0003125178960000077
means not exceeding
Figure BDA0003125178960000078
Is the largest integer of (a) to (b),
Figure BDA0003125178960000079
to represent
Figure BDA00031251789600000710
The changed value; s (p)' represents the changed value of s (p), and the expression is:
Figure BDA00031251789600000711
selecting secret key T > | SmaxWherein S represents the overall difference statistics of the sub-audio of the high frequency embedded domain, and the amount of change B of S (p) is calculated as:
Figure BDA00031251789600000712
(303) reversible embedding of watermark sequences into high frequency embedding domain A by changing S (p) histogram of moving difference statisticshIn the method, 1 bit watermark information is embedded in the sub-audio of each high-frequency embedded domain, and w (p) represents AhThe watermark embedded in the sub audio of the pth high-frequency embedding domain, w (p) is E, and E represents a watermark error; when the watermark is 0, s (p) is unchanged, and when the watermark is 1, s (p)' ═ s (p) + B, i.e.
Figure BDA00031251789600000713
The difference statistic S ═ { S (p) |1 ≦ p ≦ N }, where N is the high-frequency embedding region AhThe number of neutron audios; order to
Figure BDA0003125178960000081
Wherein α (k) represents
Figure BDA0003125178960000082
By the q-th sampling point of the k-th sampling point group in the sub-audio of the p-th high-frequency embedded domain
Figure BDA0003125178960000083
Implementing integer transforms
Figure BDA0003125178960000084
To effect a change in s (p); the integer transform expression is:
Figure BDA0003125178960000085
wherein
Figure BDA0003125178960000086
Is that
Figure BDA0003125178960000087
The value after integer transformation is composed of
Figure BDA0003125178960000088
The constructed set represents the new high frequency embedded domain
Figure BDA0003125178960000089
(304) After transformation
Figure BDA00031251789600000810
The audio may not be in the range of the original sampling point value, and the audio needs to be preprocessed before the watermark is embedded, so as to prevent the overflow phenomenon from being generated, which is obtained by the following formula:
Figure BDA00031251789600000811
since k is more than or equal to 1 and less than or equal to M,
Figure BDA00031251789600000812
therefore, the method comprises the following steps:
Figure BDA00031251789600000813
order to
Figure BDA00031251789600000814
Wherein σ represents
Figure BDA00031251789600000815
The maximum value can be obtained by firstly traversing the sampling points with embedded watermarks, marking sampling values which are not in the range of the original sampling point values, then adjusting the marked sampling point values into the original sampling point values, adjusting the values which are lower than the lower limit of the original sampling point values into the lower limit, adjusting the values which are higher than the upper limit of the original sampling point values into the upper limit, and adjusting the values which are lower than the lower limit of the original sampling point values into the upper limitThe range is large and σ is small, the overflow sampling points are few, and the impact on audio quality is small. The range of sample point values is large and σ is small, the overflow sample points are few, and the influence on the audio quality is small.
(4) Embedding new low frequency into domain
Figure BDA00031251789600000816
And a new high frequency embedded domain
Figure BDA00031251789600000817
Inverse transformation F by frequency domain transformation function-1Generating watermark-containing Audio Aw: embedding the new low frequency in step 2 into the domain
Figure BDA00031251789600000818
And the new high frequency embedded domain in step 3
Figure BDA00031251789600000819
Inverse transformation F by a transformation function-1Reconstructing watermark-containing Audio Aw。
(5) For watermark-containing audio AwExtraction of watermark information is performed as shown in fig. 2.
(501) Watermark-containing audio A through frequency domain function FwDecomposition into two independent embedded domains, including a new low frequency embedded domain
Figure BDA0003125178960000091
And a new high frequency embedded domain
Figure BDA0003125178960000092
(502) Reversible watermark information w2 is extracted by shifting the differential histogram and stored,
Figure BDA0003125178960000093
is recovered to be Ah；
(503) From
Figure BDA0003125178960000094
Mid-extraction of robust watermark w1, reversible watermark information w2
Figure BDA0003125178960000095
Is recovered to be Al；
(504) Inverse transformation F by frequency domain function-1A is to behAnd AlReverting to the original audio a.

Claims (7)

1. The reversible robust medical audio method based on the two-stage embedding is characterized by comprising the following steps:
(1) converting original audio into two independent embedded domains including a low-frequency embedded domain and a high-frequency embedded domain through a frequency domain transformation function;
(2) embedding a robust watermark in the low-frequency embedded domain by adopting a robust watermark algorithm to form a new low-frequency embedded domain;
(3) embedding the watermark error between the new low-frequency embedded domain and the low-frequency embedded domain into the high-frequency embedded domain to form a new high-frequency embedded domain, wherein the watermark error is used as reversible watermark information;
(4) generating the watermark-containing audio by inverse transformation of the frequency domain transformation function for the new low-frequency embedded domain and the new high-frequency embedded domain;
(5) and extracting watermark information from the audio containing the watermark.
2. The reversible robust medical audio method based on two-stage embedding of claim 1, wherein the step 1 audio conversion process is:
the original audio A is represented as a set of sample points, denoted as { a }1,a2,a3,a4,...,a2g-1,a2gDecomposing original audio A into low-frequency embedded domain A by wavelet transformlAnd a high frequency embedded domain AhAnd processing the sampling point pairs through a frequency domain transformation function F, wherein the expression is as follows:
Figure FDA0003125178950000011
wherein g represents the number of sampling pointsHalf, g is an integer, alRepresenting a low frequency embedded domain signal, ahRepresenting a high frequency embedded domain signal; low frequency embedded domain AlAnd a high frequency embedded domain AhThe lengths are the same.
3. The reversible robust medical audio method based on two-stage embedding of claim 2, wherein the step 2 comprises the following steps in the process of embedding the robust watermark:
(201) embedding low frequencies into domain AlUniformly dividing the audio into sub-audios which are not overlapped and have the same length, wherein each sub-audio has 2n sampling points, n represents half of the number of the sampling points of one sub-audio, n is an integer, and the sampling point of each sub-audio is marked as al(1),al(2),......,al(2n), in order to ensure information security, defining a random mapping relation:
Figure FDA0003125178950000012
the positions of the original sampling points are orderly disturbed,
Figure FDA0003125178950000013
a sequence of positions representing the sample points after mapping;
definition of
Figure FDA0003125178950000014
Representing the low frequency embedded domain AlIn the sub-audio of
Figure FDA0003125178950000015
I is more than or equal to 1 and less than or equal to 2 n;
(202) embedding a 1-bit watermark w1 for each sub-audio of the low-frequency embedded domain, wherein w1 belongs to {0,1}, and the embedding formula is as follows:
Figure FDA0003125178950000016
wherein
Figure FDA0003125178950000021
Representing the low-frequency embedded domain A after embedding a watermarklIn the sub-audio of
Figure FDA0003125178950000022
A sampling point value of
Figure FDA0003125178950000023
The composed set is represented as a new low frequency embedded domain
Figure FDA0003125178950000024
μ is watermark embedding strength, dlRepresenting the low frequency embedded domain AlThe difference value of the sampling point sets of the front part and the rear part of the sub audio is expressed as follows:
Figure FDA0003125178950000025
after embedding the robust watermark, the new low frequency embedded domain
Figure FDA0003125178950000026
The difference between the sampling points of the front and the rear parts of the sub-audio
Figure FDA0003125178950000027
The modification is as follows:
Figure FDA0003125178950000028
(203) and (4) marking the sequence of the sampling points which overflow after the change because the sampling point values overflow due to the change, restoring the sequence to the original sampling point values, and reserving new sampling point values for the sampling points which do not overflow.
4. The reversible robust medical audio method based on two-stage embedding of claim 3, wherein the watermark error in step 3 is denoted by E, and the watermark error is generatedFactors include watermark embedding strength mu and low frequency embedding domain AlThe difference value set D ═ D of sampling point sets of the front part and the back part of the sub-audiol(1),dl(2) ,.. }, wherein dl(1),dl(2) Respectively representing the low frequency embedded domain AlD of the first and second sub-audiolThe expression is:
Figure FDA0003125178950000029
wherein
Figure FDA00031251789500000210
Representing an additive binary operation.
5. The reversible robust medical audio method based on two-stage embedding of claim 4, wherein the step 3 embedding process comprises:
(301) embedding high frequencies into domain AhUniformly dividing the audio into sub-audios which are not overlapped and have equal length, wherein sampling points in each sub-audio are divided into M groups, and M represents the number of groups in one sub-audio in a high-frequency embedded domain;
each set comprising 2 sampling points per group,
Figure FDA00031251789500000211
a qth sampling point representing the kth sampling point group in the sub-audio of the pth high frequency embedding domain, q ∈ {1,2 };
Figure FDA00031251789500000212
representing the difference of the sampling points of the kth sampling point group of the sub-audio of the pth high-frequency embedded domain, k is an integer, and the expression is as follows:
Figure FDA00031251789500000213
the difference statistic of the sub-audio of the pth high-frequency embedded domain is called S (p), and the expression:
Figure FDA00031251789500000214
(302) by changing
Figure FDA0003125178950000031
And (3) realizing the change of S (p), so that the watermark is embedded reversibly, wherein the embedding expression is as follows:
Figure FDA0003125178950000032
wherein B represents the amount of change in S (p),
Figure FDA0003125178950000033
means not exceeding
Figure FDA0003125178950000034
Is the largest integer of (a) to (b),
Figure FDA0003125178950000035
to represent
Figure FDA0003125178950000036
The changed value; s (p)' represents the changed value of s (p), and the expression is:
Figure FDA0003125178950000037
selecting secret key T > | SmaxWherein S represents the overall difference statistics of the sub-audio of the high frequency embedded domain, and the amount of change B of S (p) is calculated as:
Figure FDA0003125178950000038
(303) reversible embedding of watermark sequences into high frequency embedding domain A by changing S (p) histogram of moving difference statisticshIn the method, 1 bit watermark information is embedded in the sub-audio of each high-frequency embedded domain, and w (p) represents AhThe watermark embedded in the sub audio of the pth high-frequency embedding domain, w (p) is E, and E represents a watermark error; when the watermark is 0, s (p) is unchanged, and when the watermark is 1, s (p)' ═ s (p) + B, i.e.
Figure FDA0003125178950000039
The difference statistic S ═ { S (p) |1 ≦ p ≦ N }, where N is the high-frequency embedding region AhThe number of neutron audios; order to
Figure FDA00031251789500000310
Wherein α (k) represents
Figure FDA00031251789500000311
By the q-th sampling point of the k-th sampling point group in the sub-audio of the p-th high-frequency embedded domain
Figure FDA00031251789500000312
Implementing integer transforms
Figure FDA00031251789500000313
To effect a change in s (p); the integer transform expression is:
Figure FDA00031251789500000314
wherein
Figure FDA00031251789500000315
Is that
Figure FDA00031251789500000316
The value after integer transformation is composed of
Figure FDA00031251789500000317
The constructed set represents the new high frequency embedded domain
Figure FDA00031251789500000318
(304) After transformation
Figure FDA00031251789500000319
The audio may not be in the range of the original sampling point value, and the audio needs to be preprocessed before the watermark is embedded, so as to prevent the overflow phenomenon from being generated, which is obtained by the following formula:
Figure FDA0003125178950000041
since k is more than or equal to 1 and less than or equal to M,
Figure FDA0003125178950000042
therefore, the method comprises the following steps:
Figure FDA0003125178950000043
order to
Figure FDA0003125178950000044
Wherein σ represents
Figure FDA0003125178950000045
The maximum value can be obtained by firstly passing through the sampling point with the embedded watermark and marking the sampling point value which is not at the original sampling point valueAnd (4) adjusting the marked sampling point values to be in the original sampling point values, adjusting the values which are lower than the lower limit of the original sampling point values to be in the lower limit, and adjusting the values which are higher than the upper limit of the original sampling point values to be in the upper limit.
6. The reversible robust medical audio method based on two-stage embedding of claim 5, wherein the step 4 of generating the watermarked audio is as follows: embedding the new low frequency in step 2 into the domain
Figure FDA0003125178950000046
And the new high frequency embedded domain in step 3
Figure FDA0003125178950000047
Inverse transformation F by a transformation function-1Reconstructing watermark-containing Audio Aw。
7. The reversible robust medical audio method based on two-stage embedding of claim 6, wherein the step 5 extraction process comprises:
(501) watermark-containing audio A through frequency domain function FwDecomposition into two independent embedded domains, including a new low frequency embedded domain
Figure FDA0003125178950000048
And a new high frequency embedded domain
Figure FDA0003125178950000049
(502) Reversible watermark information w2 is extracted by shifting the differential histogram and stored,
Figure FDA00031251789500000410
is recovered to be Ah；
(503) From
Figure FDA00031251789500000411
Extracting robust watermark w1 from the reversible watermark informationw2 will
Figure FDA00031251789500000412
Is recovered to be Al；
(504) Inverse transformation F by a transformation function-1A is to behAnd AlReverting to the original audio a.
CN202110687228.2A 2021-06-21 2021-06-21 Reversible robust medical audio method based on two-stage embedding Active CN113470666B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110687228.2A CN113470666B (en) 2021-06-21 2021-06-21 Reversible robust medical audio method based on two-stage embedding

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110687228.2A CN113470666B (en) 2021-06-21 2021-06-21 Reversible robust medical audio method based on two-stage embedding

Publications (2)

Publication Number Publication Date
CN113470666A true CN113470666A (en) 2021-10-01
CN113470666B CN113470666B (en) 2023-05-16

Family

ID=77868930

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110687228.2A Active CN113470666B (en) 2021-06-21 2021-06-21 Reversible robust medical audio method based on two-stage embedding

Country Status (1)

Country Link
CN (1) CN113470666B (en)

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050074139A1 (en) * 2003-10-02 2005-04-07 Seo Yong Seok Method for embedding and extracting digital watermark on lowest wavelet subband
US20080137749A1 (en) * 2001-09-10 2008-06-12 Jun Tian Assessing Quality of Service Using Digital Watermark Information
US20080215333A1 (en) * 1996-08-30 2008-09-04 Ahmed Tewfik Embedding Data in Audio and Detecting Embedded Data in Audio
CN102074237A (en) * 2010-11-30 2011-05-25 辽宁师范大学 Digital audio watermarking method based on invariant characteristic of histogram
CN104795071A (en) * 2015-04-18 2015-07-22 广东石油化工学院 Blind audio watermark embedding and watermark extraction processing method
CN108682425A (en) * 2018-05-11 2018-10-19 复旦大学 A kind of robust digital audio watermark embedded system based on constant watermark
CN109344578A (en) * 2018-10-10 2019-02-15 西安邮电大学 Based on the insertion of the audio frequency watermark of chaos and wavelet transformation, extracting method
CN110163787A (en) * 2019-04-26 2019-08-23 江苏信实云安全技术有限公司 Digital audio Robust Blind Watermarking Scheme embedding grammar based on dual-tree complex wavelet transform
CN111737662A (en) * 2020-06-16 2020-10-02 北京达佳互联信息技术有限公司 Watermark adding method, watermark extracting method, watermark adding device, watermark extracting device and electronic equipment

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080215333A1 (en) * 1996-08-30 2008-09-04 Ahmed Tewfik Embedding Data in Audio and Detecting Embedded Data in Audio
US20080137749A1 (en) * 2001-09-10 2008-06-12 Jun Tian Assessing Quality of Service Using Digital Watermark Information
US20050074139A1 (en) * 2003-10-02 2005-04-07 Seo Yong Seok Method for embedding and extracting digital watermark on lowest wavelet subband
CN102074237A (en) * 2010-11-30 2011-05-25 辽宁师范大学 Digital audio watermarking method based on invariant characteristic of histogram
CN104795071A (en) * 2015-04-18 2015-07-22 广东石油化工学院 Blind audio watermark embedding and watermark extraction processing method
CN108682425A (en) * 2018-05-11 2018-10-19 复旦大学 A kind of robust digital audio watermark embedded system based on constant watermark
CN109344578A (en) * 2018-10-10 2019-02-15 西安邮电大学 Based on the insertion of the audio frequency watermark of chaos and wavelet transformation, extracting method
CN110163787A (en) * 2019-04-26 2019-08-23 江苏信实云安全技术有限公司 Digital audio Robust Blind Watermarking Scheme embedding grammar based on dual-tree complex wavelet transform
CN111737662A (en) * 2020-06-16 2020-10-02 北京达佳互联信息技术有限公司 Watermark adding method, watermark extracting method, watermark adding device, watermark extracting device and electronic equipment

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
ALI AKBAR ATTARI: "Robust and Transparent Audio Watermarking based on Spread Spectrum in Wavelet Domain" *
陈亮: "基于扩频和回声的音频信息隐藏方法研究", 《中国优秀硕士学位论文全文数据库》 *

Also Published As

Publication number Publication date
CN113470666B (en) 2023-05-16

Similar Documents

Publication Publication Date Title
Zhang et al. Robust Reversible Audio Watermarking Scheme for Telemedicine and Privacy Protection.
Kashyap et al. Image watermarking using 3-level discrete wavelet transform (DWT)
Ramanjaneyulu et al. Wavelet-based oblivious image watermarking scheme using genetic algorithm
Malonia et al. Digital image watermarking using discrete wavelet transform and arithmetic progression technique
CN105741844B (en) A kind of digital audio watermarking algorithm based on DWT-SVD-ICA
Shahadi et al. High capacity and inaudibility audio steganography scheme
CN116456037B (en) Diffusion model-based generated image steganography method
Khare et al. Digital image watermarking scheme in wavelet domain using chaotic encryption
CN106097236B (en) Frequency domain robust image reversible water mark method based on Non-negative Matrix Factorization
CN106875324A (en) Lossless image information concealing method based on SBDE
CN103577730A (en) Reversible database watermark embedding and extracting method based on integral wavelet transformation
Bedi et al. 2L-DWTS—Steganography technique based on second level DWT
CN113470666A (en) Reversible robust medical audio method based on two-stage embedding
CN114974269A (en) Audio watermark embedding and extracting method based on singular value decomposition
Singh et al. Digital image watermarking techniques: A survey
CN110047495B (en) High-capacity audio watermarking algorithm based on 2-level singular value decomposition
CN111340675B (en) Sparse representation-based color pattern watermark embedding and extracting method
Sharma et al. Robust image watermarking technique using contourlet transform and optimized edge detection algorithm
CN110010142B (en) Large-capacity audio information hiding method
CN114255767A (en) Audio digital watermarking technology based on cross-media perception
Lee et al. Genetic algorithm-based watermarking in discrete wavelet transform domain
Behravan et al. Introducing a new method of image reconstruction against crop attack using sudoku watermarking algorithm
Padmapriya et al. Hardware Software Co-simulation of Watermarking algorithm for Image Processing Applications
CN115798490B (en) Audio watermark implantation method and device based on SIFT transformation
Alsaif et al. Contourlet transformation for text hiding in hsv color image

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant