CN107564536A - AMR pitch delay steganalysis methods based on difference Markov transition probability features in pitch delay subframe group group - Google Patents
AMR pitch delay steganalysis methods based on difference Markov transition probability features in pitch delay subframe group group Download PDFInfo
- Publication number
- CN107564536A CN107564536A CN201710797602.8A CN201710797602A CN107564536A CN 107564536 A CN107564536 A CN 107564536A CN 201710797602 A CN201710797602 A CN 201710797602A CN 107564536 A CN107564536 A CN 107564536A
- Authority
- CN
- China
- Prior art keywords
- group
- subframe
- difference
- pitch delay
- amr
- 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.)
- Pending
Links
Landscapes
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
The invention discloses the AMR pitch delay steganalysis methods based on difference Markov transition probability features in pitch delay subframe group group.This method is directed to the steganographic algorithm of two kinds of modification pitch delays in voice coding, it is proposed the AMR steganalysis features based on difference Markov transition probabilities in pitch delay subframe group group, classification prediction is carried out using SVMs, realizes the steganalysis method towards AMR pitch delays.It is an advantage of the present invention that for existing two kinds of steganography methods towards pitch delay, when relatively embedded rate is 50%, verification and measurement ratio can reach more than 95%.
Description
Technical field
It is more particularly to a kind of to judge AMR voices whether by secret letter the present invention relates to digital media processing technical field
Cease the multimedia content security technology area of steganography.
Technical background
Since in recent years, along with the continuous maturation of mobile communication means, communication manufacturer has been popularized in an all-round way for movement
The mechanics of communication means of internet, speech communication business demand amount is growing, to the coding quality and network of speech communication
The requirement General Promotion of bandwidth usage.To meet the needs of mobile voice communication, in October, 1999,3GPP (3rd
Generation Partnership Project) specify AMR (AdaptiveMulti-Rate, adaptive multi-beam forming) voice coder
Speech Coding Standards of the code as mobile Internet speech communication so that AMR is encoded in GSM, TDMA, UMTS and VoLTE
In be widely used.At present, all kinds of mobile terminal devices, such as：AMR is arranged to hand by the cell phone manufacturers such as Samsung, apple and Huawei
The phonetic matrix of machine.Meanwhile installed in the various communication APP of intelligent terminal, such as：Wechat, QQ and Skype etc., in voice-enabled chat
In tone information function, form is also encoded using AMR.With the popularization of AMR coding standards, AMR voices provide for Steganography
Brand-new space.
Being engendered for the steganographic algorithm of AMR voices, these algorithms are mainly integrated in the cataloged procedure of AMR voices,
By adjusting and changing relevant parameter in an encoding process, make to complete insertion containing secret information in parameter.Based on AMR voices
Encoding characteristics, existing steganographic algorithm relate generally to linear prediction stage in voice compression coding, adaptive codebook search rank
Section and fixed codebook stage of modulating.Wherein, for the steganographic algorithm in AMR voice adaptive codebook search stages, to cataloged procedure
Middle pitch delay, which is finely adjusted, realizes hiding for secret information, and detection difficulty is big, and corresponding steganalysis algorithm is also less.
In the AMR voice coding adaptive codebook search stages, pitch delay is the prediction result to the voice fundamental cycle, turbid
There is short-time stability between segment pitch delay.And the existing steganography method towards pitch delay is by controlling pitch delay
Realize the insertion of secret information in hunting zone, it will cause the stability characteristic (quality) in short-term of voiced segments pitch delay to change.It is based on
Above-mentioned consideration, as feature, it can be achieved to be directed to AMR bases using based on difference Markov transition probabilities in pitch delay subframe group group
Sound postpones the effective detection of steganography method.
The steganalysis detection method of the present invention is based on AMR voices, is needed before content of the invention statement to AMR voices
Encoding and decoding principle is introduced.
AMR voice compression codings are the mixing voice compression coding technologies based on ACELP, and being characterized in can be according to waiting to press
The concrete condition and communication environment of contracting audio, select suitable voice compression coding code check.Fig. 1 is AMR coding principle schematic diagrames,
When its general principle is to raw tone framing code, according to the weighted mean square error minimum for making synthesis voice and raw tone
Criterion, suitable codebook vectors are picked out from adaptive codebook and fixed codebook to replace residual signals, and by code vector address
Be sent to receiving terminal after the parameter quantization encoding of gain and each wave filter.AMR encoders input voice sample frequency be
8kHz, voice signal are that 16bit quantifies the linear PCM coding without compression, are encoded by a frame of 20ms, a frame includes 160
Individual sampled point, often it is divided into 4 subframes in units of 5ms in frame.Three parts can be divided mainly into according to the function of its realization, wrap
Include linear prediction stage, adaptive codebook search stage and fixed codebook stage of modulating.Technical term solution to be related to below
Release：
1st, audio is compressed：Refer to the audio Jing Guo lossy compression method, if MP3, WMA, AMR are lossy compression method audio.
2、Cover：Carrier audio, i.e., the audio of secret information insertion is not carried out.
3、Stego：Close audio is carried, that is, has carried out the audio of secret information insertion.
4th, subframe group：AMR voices include 4 subframes per frame, the first subframe and the second subframe one subframe group of composition, and the 3rd
Subframe and the 4th subframe form a subframe group.
5th, relation in group：Represent the relation between previous subframe pitch delay and latter subframe pitch delay in subframe group.
The content of the invention
The problem of present invention for lacking relatively towards AMR steganalysis methods at present, realize towards AMR pitch delays
Steganalysis.
Technical scheme is used based on difference Markov transition probabilities in pitch delay subframe group group as in short-term
Estimation of stability standard extracts feature, and two classification are carried out to Cover the and Stego samples of AMR voices using SVM classifier.1、
A kind of AMR pitch delay steganalysis methods based on difference Markov transition probability features in pitch delay subframe group group, its
It is characterised by, including：
Step 1, based on difference Markov transition probabilities in pitch delay subframe group group as short-time stability evaluation criterion
Feature is extracted, is specifically included：
Difference in step 1.1, structure pitch delay subframe group group, in AMR voice compression codings, 20ms voice is one
Frame, a frame voice are that unit is divided into 4 subframes by 5ms, and each subframe has a pitch delay；In a frame speech frame, the
Two subframe T2With the 4th subframe T4Pitch delay respectively with the first subframe T1With the 3rd subframe T3Based on, in certain section
Obtained according to correlation；As shown in Fig. 2 by speech frame, the first subframe T1With the second subframe T2As a subframe group, the 3rd
Subframe T3With the 4th subframe T4As another subframe group；
The pitch delay sequence of AMR voices is expressed as P=(p in the way of subframe group for ease of description11,p12,...,
pt1,pt2,...,pN1,pN2), N is the total number of subframe group in AMR voices, and i is the chronological index value of subframe group, pt1
And pt2First subframe and second subframe in respectively subframe group t；Because voice voiced segments pitch delay is with steady in short-term
It is qualitative, calculate difference in subframe group group and be used to measure the stability in subframe group group between two subframes；Difference in subframe group group
DintraCalculation formula is as shown in Equation 1, t ∈ [1, N]：
Dintra(t)=pt1-pt2Formula 1
Difference Markov transition probability features in step 1.2, structure pitch delay subframe group group：Markov transition probabilities
The effect of matrix is to weigh the probability that variable is changed between different states, therefore can use it to and represent continuous base
The change situation of difference in sound delay subframe group group；Difference Markov transition probabilities are as hidden in calculating pitch delay subframe group group
Analysis feature is write, the difference condition of difference in Cover voices and Stego voice fundamentals delay subframe group group is described, realizes to two
The differentiation of person；
Difference Markov transition probabilities characterizing definition is M1 in pitch delay subframe group groupintra, calculated by formula 2：
M1intraThe value of (i, j) is difference D in groupintra(t) when being i, Dintra(t+1) Markov transition probabilities for being j, t are in group
Index in difference sequential, S0For the total number of difference in the subframe group group that is calculated
Understood according to AMR codings, in a frame voice, the second subframe T2With the 4th subframe T4Pitch delay respectively with first
Subframe T1With the 3rd subframe T3Based on, obtained in certain section according to correlation；Therefore, for different code checks, internal difference is organized
The section of value is fixed, difference D in groupintraScope, and corresponding M1intraCharacteristic dimension, by extracting AMR voices
Difference Markov transition probabilities in pitch delay subframe group group, as the steganography characteristic of division changed pitch delay；
Step 2, two classification are carried out to Cover the and Stego samples of AMR voices using SVM classifier and carry out steganography point
Analysis detection.
In a kind of above-mentioned AMR pitch delays based on difference Markov transition probability features in pitch delay subframe group group
Steganalysis method, the step 2 specifically include：
Step 2.1, classifier training, are specifically included：
Step 2.1.1, WAV samples are inputted, generate cover samples and corresponding stego samples respectively, and according in 1
Difference Markov transition probability characteristic of divisions in method extraction pitch delay subframe group group；
Step 2.1.2, after 2.1.1 processes, obtain training set sample and the equal two kinds of different embedded mobile GISs of quantity
The close sample of load, the stego samples and cover samples for then randomly choosing varying number use SVM classifier training steganography point
Analyse model；
Step 2.2, steganalysis detection, are specifically included：
The flow that steganalysis detection is carried out using above-mentioned steganalysis model is comprised the following steps：
Step 2.2.1, extract the steganalysis characteristic set of sample to be tested；
Step 2.2.2, the steganalysis model that feature input is built, obtains the steganography judged result of sample.
Brief description of the drawings
Fig. 1 is AMR encoding and decoding flow charts.
Fig. 2 is difference schematic diagram in the division of subframe group and group.
Fig. 3 a are carrier audios with carrying distribution of the difference in close audio pitch delay distribution and pitch delay subframe group group
(Cover voices and Stego voice fundamentals delay distribution).
Fig. 3 b are carrier audios with carrying distribution of the difference in close audio pitch delay distribution and pitch delay subframe group group
(distribution of the difference situation in Cover voices and Stego voice fundamentals delay subframe group group).
Fig. 4 a are difference Markov transition probability features in pitch delay subframe group group of the carrier audio with carrying close audio
(Cover voices).
Fig. 4 b are difference Markov transition probability features in pitch delay subframe group group of the carrier audio with carrying close audio
(Stego voices).
Fig. 5 is steganalysis training and the detection framework of the present invention.
Embodiment
1st, feature extracting method
1.1 feature extraction
1.1.1 difference in pitch delay subframe group group is built
In AMR voice compression codings, 20ms voice is a frame, and a frame voice is that unit is divided into 4 subframes by 5ms,
Each subframe has a pitch delay.In a frame speech frame, the second subframe T2With the 4th subframe T4Pitch delay respectively with
First subframe T1With the 3rd subframe T3Based on, obtained in certain section according to correlation.As shown in Fig. 2 by speech frame,
First subframe T1With the second subframe T2As a subframe group, the 3rd subframe T3With the 4th subframe T4As another subframe group.
The pitch delay sequence of AMR voices is expressed as P=(p in the way of subframe group for ease of description11,p12,...,
pt1,pt2,...,pN1,pN2), N is the total number of subframe group in AMR voices, and i is the chronological index value of subframe group, pt1
And pt2First subframe and second subframe in respectively subframe group t.Because voice voiced segments pitch delay is with steady in short-term
It is qualitative, calculate difference in subframe group group and be used to measure the stability in subframe group group between two subframes.Difference in subframe group group
DintraCalculation formula is as shown in Equation 1, t ∈ [1, N]：
Dintra(t)=pt1-pt2(formula 1)
1.1.2 difference Markov transition probability features in pitch delay subframe group group are built
The effect of Markov transition probabilities matrix is to weigh the probability that variable is changed between different states, because
This can use it to the change situation for representing difference in successive pitch delay subframe group group.Calculate in pitch delay subframe group group
Difference Markov transition probabilities are described in Cover voices and Stego voice fundamentals delay subframe group group as steganalysis feature
The difference condition of difference, realize the differentiation to both.
Difference Markov transition probabilities characterizing definition is M1 in pitch delay subframe group groupintra, calculated by formula 2：
M1intraThe value of (i, j) is difference D in groupintra(t) when being i, Dintra(t+1) Markov transition probabilities for being j, t are in group
Index in difference sequential, S0For the total number of difference in the subframe group group that is calculated
Understood according to AMR codings, in a frame voice, the second subframe T2With the 4th subframe T4Pitch delay respectively with first
Subframe T1With the 3rd subframe T3Based on, obtained in certain section according to correlation.Therefore, for different code checks, internal difference is organized
The section of value is fixed, difference D in groupintraScope, and corresponding M1intraShown in characteristic dimension table 1.
Difference D in 1 group of tableintraScope, M1intraCharacteristic dimension
Postpone difference Markov transition probabilities in subframe group group by extracting AMR voice fundamentals, repaiied as to pitch delay
The steganography characteristic of division changed.
1.2 characteristic principles are analyzed
Because the pitch period of voice voiced segments has short-time stability, pitch delay is to fundamental tone in speech
The prediction result in cycle, therefore voiced segments pitch delay also has short-time stability, should be between adjacent pitch delay compared with
For stabilization, the difference of two pitch delays should be smaller in subframe group.The existing steganography method for pitch delay, pass through
Steganography is realized in the hunting zone of pitch delay in control cataloged procedure.The pitch lag values that steganography process causes search to obtain occur
Change, the short-time stability between the adjacent pitch delay of voiced segments can be destroyed, the difference meeting of two pitch delays in subframe group
Become big.
Fig. 3 is it is shown that one section of voiced segments pitch delay sequence of Cover voices and Stego voices, subframe group group internal difference
Value DintraDistribution situation.Wherein Cover voices are the AMR voices texts of open AMR encoders generation 12.2kbps bit rate modes
Part, Stego voices are the AMR speech samples generated after being embedded in using document [1] steganographic algorithm by maximum capacity.Fig. 3 (a) tables
Bright, the pitch delay between Cover speech subframes is more steady than Stego voice.Fig. 3 (b) is in Cover voices and Stego voices
Difference D in subframe group groupintraDistribution situation, it can be seen that both have a significant difference, in Cover voices, the small distribution of difference
To concentrate, continuous difference change is small, the phenomenon that continuous difference is 0 be present, and in Stego voices, the value of distribution of the difference is larger, continuously
Difference changes greatly, and difference is 0 situation very little.Comparing result shows that steganography process destroys the adjacent fundamental tone of voiced segments and prolonged really
Short-time stability between late.
The effect of Markov transition probabilities matrix is to weigh the probability that variable is changed between different states, because
This can use it to the change situation represented between continuous difference, calculate difference Markov transfers in pitch delay subframe group group
Probability describes the difference condition of Cover voices and Stego voices, it is possible to achieve Cover voices and Stego voices as feature
Differentiation.
To Cover voices and Stego voices, difference Markov transition probabilities point in pitch delay subframe group group are calculated respectively
Cloth situation is shown in Fig. 4.Wherein Cover voices are the AMR voices texts of open AMR encoders generation 12.2kbps bit rate modes
Part, Stego voices are the AMR speech samples generated after being embedded in using document [1] steganographic algorithm by maximum capacity.Can from result
To find out, in Cover voices, there is obvious peak value in feature distribution in core, and in Stego voices feature distribution compared with
To be flat, therefore difference Markov transition probabilities are present in Cover voices and in Stego voices in pitch delay subframe group group
Significant difference.
It can be seen that from Fig. 3, Fig. 4 comparing result difference Markov transition probabilities in pitch delay subframe group group is special
Sign is as characteristic of division come to distinguish cover audios with stego audios be effective.
2nd, steganalysis detects
2.1 classifier training
Step 2.1.1, WAV samples are inputted, generate cover samples and corresponding stego samples respectively, and according in 1
Difference Markov transition probability characteristic of divisions in method extraction pitch delay subframe group group.
Step 2.1.2, after 2.1.1 processes, obtain training set sample and the equal two kinds of different embedded mobile GISs of quantity
The close sample of load, the stego samples and cover samples for then randomly choosing varying number use SVM classifier training steganography point
Analyse model.
2.2 steganalysis detect
The flow that steganalysis detection is carried out using above-mentioned steganalysis model is comprised the following steps：
Step 2.2.1, extract the steganalysis characteristic set of sample to be tested.
Step 2.2.2, the steganalysis model that feature input is built, obtains the steganography judged result of sample.
2.3 steganalysis experimental results
In order to verify the validity of inventive algorithm, the present invention is directed to the steganalysis model of different steganography methods training,
Experimental result is as shown in table 1, table 2.Wherein TPR table shows the probability for being detected as Stego for carrying close audio (Stego), and TNR represents to carry
Body audio (Cover) is detected as Cover verification and measurement ratio.
Test result indicates that in 12.2kbps, 10.2kbps, 7.95kbps, 7.40kbps, 6.70kbps and 5.90kbps
Under coding mode, disaggregated model of the invention to existing two kinds for scalefactor bands code book modification steganography methods have compared with
Good detectability, when relatively embedded rate is 50%, verification and measurement ratio is attained by more than 95%.
Table 2 detects the performance of document [1] steganographic algorithm
Table 3 detects the performance of document [2] steganographic algorithm
Specific embodiment described herein is only to spirit explanation for example of the invention.Technology belonging to the present invention is led
The technical staff in domain can be made various modifications or supplement to described specific embodiment or be replaced using similar mode
Generation, but without departing from the spiritual of the present invention or surmount scope defined in appended claims.
Claims (2)
- A kind of 1. AMR pitch delays steganalysis side based on difference Markov transition probability features in pitch delay subframe group group Method, it is characterised in that including：Step 1, based on difference Markov transition probabilities in pitch delay subframe group group as short-time stability evaluation criterion extract Feature, specifically include：Difference in step 1.1, structure pitch delay subframe group group, in AMR voice compression codings, 20ms voice is a frame, One frame voice is that unit is divided into 4 subframes by 5ms, and each subframe has a pitch delay；In a frame speech frame, the second son Frame T2With the 4th subframe T4Pitch delay respectively with the first subframe T1With the 3rd subframe T3Based on, in certain section according to Correlation obtains；As shown in Fig. 2 by speech frame, the first subframe T1With the second subframe T2As a subframe group, the 3rd subframe T3With the 4th subframe T4As another subframe group；The pitch delay sequence of AMR voices is expressed as P=(p in the way of subframe group for ease of description11,p12,...,pt1, pt2,...,pN1,pN2), N is the total number of subframe group in AMR voices, and i is the chronological index value of subframe group, pt1And pt2 First subframe and second subframe in respectively subframe group t；Because voice voiced segments pitch delay has short-time stability, Difference in subframe group group is calculated to be used to measure the stability in subframe group group between two subframes；Difference D in subframe group groupintraMeter It is as shown in Equation 1 to calculate formula, t ∈ [1, N]：Dintra(t)=pt1-pt2Formula 1Difference Markov transition probability features in step 1.2, structure pitch delay subframe group group：Markov transition probabilities matrix Effect be to weigh the probability that is changed between different states of variable, therefore can use it to and represent that successive pitch prolongs The change situation of difference in slow subframe group group；Difference Markov transition probabilities in pitch delay subframe group group are calculated as steganography to divide Feature is analysed, the difference condition of difference in Cover voices and Stego voice fundamentals delay subframe group group is described, realizes to both Distinguish；Difference Markov transition probabilities characterizing definition is M1 in pitch delay subframe group groupintra, calculated by formula 2：M1intra The value of (i, j) is difference D in groupintra(t) when being i, Dintra(t+1) Markov transition probabilities for being j, when t is difference in group Index in sequence, S0For the total number of difference in the subframe group group that is calculatedUnderstood according to AMR codings, in a frame voice, the second subframe T2With the 4th subframe T4Pitch delay respectively with the first subframe T1With the 3rd subframe T3Based on, obtained in certain section according to correlation；Therefore, for different code checks, difference in group Section is fixed, difference D in groupintraScope, and corresponding M1intraCharacteristic dimension, by extracting AMR voice fundamentals Postpone difference Markov transition probabilities in subframe group group, as the steganography characteristic of division changed pitch delay；Step 2, two classification are carried out to Cover the and Stego samples of AMR voices using SVM classifier and carry out steganalysis inspection Survey.
- It is 2. according to claim 1 a kind of based on difference Markov transition probability features in pitch delay subframe group group AMR pitch delay steganalysis methods, it is characterised in that the step 2 specifically includes：Step 2.1, classifier training, are specifically included：Step 2.1.1, WAV samples are inputted, generate cover samples and corresponding stego samples respectively, and according to the method in 1 Extract difference Markov transition probability characteristic of divisions in pitch delay subframe group group；Step 2.1.2, after 2.1.1 processes, obtain the load of training set sample and the equal two kinds of different embedded mobile GISs of quantity Close sample, the stego samples and cover samples for then randomly choosing varying number use SVM classifier training steganalysis mould Type；Step 2.2, steganalysis detection, are specifically included：The flow that steganalysis detection is carried out using above-mentioned steganalysis model is comprised the following steps：Step 2.2.1, extract the steganalysis characteristic set of sample to be tested；Step 2.2.2, the steganalysis model that feature input is built, obtains the steganography judged result of sample.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710797602.8A CN107564536A (en) | 2017-09-06 | 2017-09-06 | AMR pitch delay steganalysis methods based on difference Markov transition probability features in pitch delay subframe group group |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710797602.8A CN107564536A (en) | 2017-09-06 | 2017-09-06 | AMR pitch delay steganalysis methods based on difference Markov transition probability features in pitch delay subframe group group |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107564536A true CN107564536A (en) | 2018-01-09 |
Family
ID=60978242
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710797602.8A Pending CN107564536A (en) | 2017-09-06 | 2017-09-06 | AMR pitch delay steganalysis methods based on difference Markov transition probability features in pitch delay subframe group group |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107564536A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110111799A (en) * | 2019-04-28 | 2019-08-09 | 武汉大学 | A kind of safe steganography method of AMR fixed codebook based on distribution of pulses model |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1003324A2 (en) * | 1995-05-08 | 2000-05-24 | Digimarc Corporation | Forgery-resistant documents with images conveying secret data and related methods |
| US20030009670A1 (en) * | 2001-04-02 | 2003-01-09 | Digimarc Corporation | Background watermark processing |
| CN101887723A (en) * | 2007-06-14 | 2010-11-17 | 华为终端有限公司 | Fine tuning method and device for pitch period |
| CN103985389A (en) * | 2014-05-30 | 2014-08-13 | 武汉大学 | Steganalysis method for AMR audio files |
| US20140340209A1 (en) * | 2013-05-17 | 2014-11-20 | Immersion Corporation | Low-frequency effects haptic conversion system |
| CN105118512A (en) * | 2015-07-23 | 2015-12-02 | 武汉大学 | General steganalysis method facing AAC digital audio |
| CN105575401A (en) * | 2015-12-23 | 2016-05-11 | 武汉大学 | AACHuffman domain steganalysis method based on C-MAC characteristics |
-
2017
- 2017-09-06 CN CN201710797602.8A patent/CN107564536A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1003324A2 (en) * | 1995-05-08 | 2000-05-24 | Digimarc Corporation | Forgery-resistant documents with images conveying secret data and related methods |
| US20030009670A1 (en) * | 2001-04-02 | 2003-01-09 | Digimarc Corporation | Background watermark processing |
| CN101887723A (en) * | 2007-06-14 | 2010-11-17 | 华为终端有限公司 | Fine tuning method and device for pitch period |
| US20140340209A1 (en) * | 2013-05-17 | 2014-11-20 | Immersion Corporation | Low-frequency effects haptic conversion system |
| CN103985389A (en) * | 2014-05-30 | 2014-08-13 | 武汉大学 | Steganalysis method for AMR audio files |
| CN105118512A (en) * | 2015-07-23 | 2015-12-02 | 武汉大学 | General steganalysis method facing AAC digital audio |
| CN105575401A (en) * | 2015-12-23 | 2016-05-11 | 武汉大学 | AACHuffman domain steganalysis method based on C-MAC characteristics |
Non-Patent Citations (2)
| Title |
|---|
| YANZHEN REN等: "AMR Steganalysis Based on Second-Order Difference of Pitch Delay", 《IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY》 * |
| 贺欣等: "基于浊音特性的语音压缩域隐写分析", 《计算机研究与发展》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110111799A (en) * | 2019-04-28 | 2019-08-09 | 武汉大学 | A kind of safe steganography method of AMR fixed codebook based on distribution of pulses model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9875752B2 (en) | Voice profile management and speech signal generation | |
| CN101131817B (en) | Method and apparatus for robust speech classification | |
| CN100362568C (en) | Method and apparatus for predictively quantizing voiced speech | |
| CN103177726B (en) | The classification of audio signal | |
| CN103548081B (en) | The sane speech decoding pattern classification of noise | |
| US11328739B2 (en) | Unvoiced voiced decision for speech processing cross reference to related applications | |
| KR20070015155A (en) | Coding model selection | |
| CN107293311A (en) | Very short pitch determination and coding | |
| US9524720B2 (en) | Systems and methods of blind bandwidth extension | |
| Gupta et al. | Speech bandwidth extension with wavenet | |
| KR20020040910A (en) | A predictive speech coder using coding scheme selection patterns to reduce sensitivity to frame errors | |
| JP2002544551A (en) | Multipulse interpolation coding of transition speech frames | |
| CN104517612A (en) | Variable-bit-rate encoder, variable-bit-rate decoder, variable-bit-rate encoding method and variable-bit-rate decoding method based on AMR (adaptive multi-rate)-NB (narrow band) voice signals | |
| García et al. | Automatic emotion recognition in compressed speech using acoustic and non-linear features | |
| Hicsonmez et al. | Audio codec identification from coded and transcoded audios | |
| CN107564536A (en) | AMR pitch delay steganalysis methods based on difference Markov transition probability features in pitch delay subframe group group | |
| Lin et al. | Mixed excitation linear prediction coding of wideband speech at 8 kbps | |
| Al-Heeti et al. | Voice encoding for wireless communication based on LPC, RPE, and CELP | |
| Albahri et al. | Artificial bandwidth extension to improve automatic emotion recognition from narrow-band coded speech | |
| Petracca et al. | Performance analysis of compressed-domain automatic speaker recognition as a function of speech coding technique and bit rate | |
| CN110097887A (en) | A kind of safe steganography method of SILK based on LSF coefficient Statistical Distribution Characteristics | |
| Huong et al. | A new vocoder based on AMR 7.4 kbit/s mode in speaker dependent coding system | |
| Li et al. | A generation method for acoustic two-dimensional barcode | |
| Jenner | Non-intrusive identification of speech codecs in digital audio signals | |
| TWI381368B (en) | Coding mode selection device |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180109 |