CN104078047B - Quantum compression method based on voice Multi-Band Excitation LSP parameters - Google Patents
Quantum compression method based on voice Multi-Band Excitation LSP parameters Download PDFInfo
- Publication number
- CN104078047B CN104078047B CN201410294051.XA CN201410294051A CN104078047B CN 104078047 B CN104078047 B CN 104078047B CN 201410294051 A CN201410294051 A CN 201410294051A CN 104078047 B CN104078047 B CN 104078047B
- Authority
- CN
- China
- Prior art keywords
- quantum
- represent
- lsp
- voice
- ket
- 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.)
- Active
Links
- 230000006835 compression Effects 0.000 title claims abstract description 34
- 238000007906 compression Methods 0.000 title claims abstract description 34
- 238000000034 method Methods 0.000 title claims abstract description 15
- 230000005284 excitation Effects 0.000 title claims abstract description 13
- 230000009466 transformation Effects 0.000 claims abstract description 9
- 238000000605 extraction Methods 0.000 claims abstract description 6
- 230000006837 decompression Effects 0.000 claims abstract description 4
- 239000013598 vector Substances 0.000 claims description 14
- 230000005540 biological transmission Effects 0.000 claims description 9
- 239000002096 quantum dot Substances 0.000 claims description 8
- 238000005259 measurement Methods 0.000 claims description 6
- 230000000295 complement effect Effects 0.000 claims description 4
- 238000004132 cross linking Methods 0.000 claims description 3
- 230000007246 mechanism Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims 1
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 229910002056 binary alloy Inorganic materials 0.000 abstract description 3
- 238000004891 communication Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008602 contraction Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Abstract
The invention discloses a kind of quantum compression method based on voice Multi-Band Excitation LSP parameters, it is related to coding techniques field.By parameter extraction, compression, sequence, it is classical binary sequence by the LSP Parameter Switch of voice, length is the classical bit of binary system 8, using unitary transformation U, list entries to different subspace enters line translation, transmitting terminal is right | and the 8 bits of ψ ' ＞ are measured, receiving terminal behind the information that is received to adding | 0 ＞, and carry out decompression operation, eight kets that will be obtained are corresponding to classical binary message, the final LSP parameters of generation, are ranked up to code book according to the parameter number of times chosen.The present invention is combined the speech model multi band excitation model of application using quantum and compress speech, reasonably using the correlation between each ket, fills up the blank of prior art, meets the use demand of people.
Description
Technical field：
The present invention relates to coding techniques field, more particularly to a kind of quantum based on voice Multi-Band Excitation LSP parameters
Compression method.
Background technology：
In classical communication field, the information of transmission has certain redundancy, removes information redundancy to reduce transmission information
The method of bit is referred to as compression, and the compression of information can play the purpose for economizing on resources.In quantum communications field, also will the amount of carrying out
Son compression, the purpose of transmission of quantum bit is saved to reach.
The method of existing quantum compressed encoding puts forward on the basis of quantum information theory, is mainly distributed formula pressure
Contracting, regular length compression, elongated compression etc..And, the specific application of quantum compression algorithm is only in terms of compression of images.Wherein：
Distributed compression refers to that source quantum state is distributed in two or more groups, and each group is sending a message to
Before recipient, the quantum state to assigning to carries out independent compression, and recipient can accordingly rebuild and be originally inputted.Such compression side
Formula, it is impossible to reasonably using the correlation between each ket, therefore only verified in theory, and without example.
, it is necessary to the density matrix of information source in regular length compression, and the characteristic value tensor space of density matrix is called allusion quotation
Type subspace, the information source information of transmission determines whether type sequence in transmitting terminal, if type sequence, then can be with less
Quantum bit sends；If non-typical sequence, any any sequence sent in typical subspace to receiving terminal.
Elongated compression is the new quantum code word that quantum code word is changed into contraction in length by unitary transformation, and by other canals
The length of commercial weight subcode word is informed receiving terminal by road, by receiving terminal according to the quantum code word and true form word length for receiving, to receiving
Code word adds a certain number of 0, code word size is recovered former long, then carry out the inverse transformation of the unitary transformation and recover commercial weight subcode
Word.
Also scholar proposes the quantum backpropagation (quantum based on quantum neural model and artificial neural network
Reverse propagation, QBP) algorithm, and for compression of images, obtained being calculated than BP with less study iterative process
The more preferable reconstructed image quality of method.
Therefore, up to the present, many people have studied quantum compression problem, but someone is not by quantum and language
The example that sound press contracting is combined.
Many band excitation (Multi-Band Excitation, MBE) models are that Griffin and Lim were directed to tradition in 1988
Shortcoming in binary model proposes a kind of new speech model.Simple binary is clear/voiced sound model in, one whole section of voice is complete
It is voiceless sound or voiced sound to adjudicate, but in actual speech, this section of voice not only containing voiceless sound but also may contain voiced sound, if according to
It is one kind that certain standard is only adjudicated, then may produce buzz in voice is synthesized.
The content of the invention：
The blank in the field, quantum of the present invention based on voice Multi-Band Excitation LSP parameters are filled up regarding to the issue above
Compression method, comprises the following steps by taking the code book of 256 sizes as an example：
1) be classical binary sequence by the LSP Parameter Switch of voice by parameter extraction, compression, sequence, length be through
The bit of allusion quotation binary system 8；
2) binary zero and 1 are corresponded into quantum state respectively | a ＞ and | b ＞ can be obtained | aaaaaaaaa ＞, |
Aaaaaaaab ＞ ... | 256 kinds of kets of bbbbbbbbb ＞；
3) according to table 1, will | it is larger just that M ＞={ | aaaaaaaa ＞, | aaaaaaab ＞ ... | bbbbbbbb ＞ } is divided into probability
Crosslinking Mechanisms
And the less orthogonal complementary subspace of probability
Wherein,Represent wherein | λa＞ and | λbThe number of the ＞ quantum ket different from preceding identical but position；
4) unitary transformation U, the list entries to different subspace is applied to enter line translation；
5) transmitting terminal is right | and the 8 bits of ψ ' ＞ are measured；
6) receiving terminal behind the information that is received to adding | 0 ＞, and decompression operation is carried out, the following two kinds shape can be obtained
One of state
7) the eight kets correspondence that will be obtained generates final LSP parameters to classical binary message.
Preferably, the LSP parameters are compressed and are transmitted in the form of code book；Each group of LSP parameter is 10 dimensions
Vector, when a frame phonetic entry need compression when, found from code book with the vectorial immediate vector, only transmit the code
This coding, if the size of code book is N, the classical number of bits for needing transmission is log2N。
Preferably, the sound bank of the self-contained China and foreign countries men and women, old and young voice of the LSP parameter extractions of the voice, uses N sizes
Code book be compressed, the selected number of times of LSP parameters in record code book, and according to choosing number of times to be ranked up；With N=256
As a example by, it is 00000000 by the LSP parameter codings for choosing number of times most, next is 00000001, is again 00000010.
Preferably, the cryptoprinciple is that the number of times chosen is more, it is more comprising 0 number in parameter coding；Conversely,
It is then most comprising 1 number.
Preferably, described right | measurement result is 0 when the 8 bits of ψ ' ＞ are measured, then Information State is |
Fghijkl0 ＞, then prove be by the quantum bit of transmission | fghijkl ＞ are typical state | ψtyp＞, uses | ψcomp1＞ represents, channel
Send | ψcomp1〉；Measurement result is 1, then Information State is | zyxwvut1 ＞, then prove that the quantum bit for sending is SARS kenel |
Zyxwvut ＞, i.e., | ψcomp2＞, sends through quantum channel | ψcomp2＞, wherein | ψcom＞ meetsTypical state.
Beneficial effect of the present invention：The present invention has filled up the blank in the field, and quantum compression method is applied into many band excitations
In model, code book is ranked up according to the number of times chosen, one section of voice is divided into smaller subband, entered in each subband
Row judgement, much closer to actual speech situation, improves the quality of synthesis voice, reasonably using the correlation between each ket
Property, meet the use demand of people.
Brief description of the drawings：
For ease of explanation, the present invention is described in detail by following specific implementations and accompanying drawing.
Fig. 1 is present invention coding flow chart；
Fig. 2 is LSP compression of parameters flow chart of the present invention.
Specific embodiment：
As shown in figure 1, this specific embodiment uses following technical scheme：Quantum pressure by taking the code book of 256 sizes as an example
The step of contracting, is as follows：
1) be classical binary sequence by the LSP Parameter Switch of voice by parameter extraction, compression, sequence, length be through
The bit of allusion quotation binary system 8；
2) binary zero and 1 are corresponded into quantum state respectively | a ＞ and | b ＞ can be obtained | aaaaaaaaa ＞, |
Aaaaaaaab ＞ ... | 256 kinds of kets of bbbbbbbbb ＞.
3) according to table 1, will | it is larger just that M ＞={ | aaaaaaaa ＞, | aaaaaaab ＞ ... | bbbbbbbb ＞ } is divided into probability
Crosslinking Mechanisms
And the less orthocomplement, orthogonal complement of probability is in space
Wherein,Represent wherein | λa＞ and | λbThe number of the ＞ quantum ket different from preceding identical but position.
4) unitary transformation U, the list entries to different subspace is applied to enter line translation：
For 163 typical states, be converted to
For 93 SARS kenels, be converted toI.e. acquired results are | ψ ' ＞=U | ψ ＞.
5) transmitting terminal：It is right | the 8 bits of ψ ' ＞ are measured
Measurement result is 0, then Information State is | fghijkl0 ＞, then prove be by the quantum bit of transmission | fghijkl ＞ are
Typical state | ψtyp＞, uses | ψcomp1＞ represents that channel sends | ψcomp1〉；
Measurement result is 1, then Information State is | zyxwvut1 ＞, then prove that the quantum bit for sending is SARS kenel |
Zyxwvut ＞, i.e., | ψcomp2＞, sends through quantum channel | ψcomp2＞, wherein | ψcom＞ meetsTypical state.
6) receiving terminal：To being added behind the information that is received | 0 ＞, and decompression operation is carried out, the following two kinds can be obtained
One of state
The eight kets correspondence that will be obtained generates final LSP parameters to classical binary message.
As shown in Figure 2, LSP parameters are compressed and are transmitted in classical communication field in the form of code book.Each group
LSP parameters are the vectors of 10 dimensions.When a frame phonetic entry need compression when, found from the code book being already prepared to this to
Immediate vector is measured, the numbering of the code book is only transmitted.If the size of code book is N, the classical bit of transmission is needed
Number is log2N。
Using the sound bank comprising China and foreign countries' men and women, old and young's voice as phonetic entry, after extracting the LSP parameters of voice in the storehouse,
It is compressed with the code book of N sizes, the selected number of times of LSP parameters in record code book, and according to choosing number of times to be ranked up.With
As a example by N=256, by the LSP parameters for choosing number of times most numbering be 00000000, next is 00000001, again for
00000010.Cryptoprinciple is that the number of times chosen is more, more comprising 0 number in parameter numbering；Conversely, then comprising 1
Number is most.
So, the LSP parameters transmitted on channel are a series of classical binary sequences.By the 0 of classical information, 1 pair
Should be Quantum Pure | a ＞, | b ＞, and carry out quantum compression.
The character density operator ρ=p of calculating informationa|a〉〈a|+pb| b ＞ ＜ b |, and obtain the characteristic vector of ρ | λa＞ and |
λb〉.The symbolic information of selection meets the condition of the characteristic vector Complete Orthogonal of ρ, therefore | λa＞ and | λb＞ Special compositions U's is complete
Orthogonal basic vector.
Characteristic vector is with overlapping for quantum state
So that codebook size is 256 as an example, then the Serial No. of code book is finally represented
00000000~11111111, corresponding to quantum state is
| M ＞={ | aaaaaaaa ＞, | aaaaaaab ＞ ... | bbbbbbbb ＞ }, totally 256 kinds of kets.
| the orthogonal basic vector { λ in M ＞ correspondences space1, λ2, λ3, λ4, λ5, λ6, λ7, λ8=λa, λb, all basic vector are obtained respectively
In order to carry out the division of typical case and atypia space, it is necessary to calculate overlapping | ＜ ψ | the M ＞ of message and orthogonal basic vector |2, it is as follows,
The like, following table can be obtained：
Quantum information length is n=8, the typical subspace of definition
Probability be 1- δ=
0.9974
Wherein,
Claims (4)
1. the quantum compression method of voice Multi-Band Excitation LSP parameters is based on, it is characterised in that the number of codebook size is N,
Wherein N=256, comprises the following steps：
1) it is classical binary sequence by the LSP Parameter Switch of voice by parameter extraction, compression, sequence, length is classical two
The bit of system 8；
2) binary zero and 1 are corresponded into quantum state respectively | a>With | b>Can obtain | aaaaaaaa>, | aaaaaaab>…
|bbbbbbbb>256 kinds of kets；
3) by this 256 kinds of kets | M>=| aaaaaaaa>, | aaaaaaab>…|bbbbbbbb>To be divided into probability larger just
Crosslinking Mechanisms
And the less orthogonal complementary subspace of probability
Wherein,Represent | λa>With | λb>The number quantum ket different from preceding identical but position, | λa>With | λb>Represent letter
Density operator ρ=the p of breatha|a><a|+pb|b><B | characteristic vector, wherein paAnd pbRespectively ket | a>With | b>Occurred
Probability；
4) unitary transformation U, the list entries to different subspace is applied to enter line translation；
5) transmitting terminal is right | ψ '>8 bits measure, | ψ '>Represent carries out unitary to the list entries of different subspace
The ket obtained after conversion, while only transmitting the first seven bit to receiving terminal；
6) receiving terminal behind the information that is received to adding | and 0>, and carry out decompression operation, can obtain the following two kinds state it
One
In the formula, | ψ ">When what expression transmitting terminal sent is typical state, the ket that receiving terminal is received；|ψ″′>Represent transmitting terminal
Send be atypia state when, receiving terminal receive ket；|ψcomp1>Represent that the typical state in orthogonal intersection space becomes by unitary
Change and delete the last position | 0>Obtained from ket；|ψcomp2>Represent the SARS kenel in orthogonal complementary subspace by unitary
Convert and delete the last position | 1>Obtained from ket；U-1Represent the inverse transformation of unitary transformation；|fghijklm0>Represent |
ψcomp1>With | 0>Carry out the ket obtained after tensor product；|ψlyp>Represent typical state；
7) the eight kets correspondence that will be obtained generates final LSP parameters to classical binary message；
Wherein, the LSP parameters are compressed and are transmitted in the form of code book；Each group of LSP parameter is the vector of 10 dimensions, when
One frame phonetic entry need compression when, found from code book with the vectorial immediate vector, only transmit the coding of the code book
.
2. the quantum compression method based on voice Multi-Band Excitation LSP parameters according to claim 1, its feature exists
In：The sound bank of the self-contained China and foreign countries men and women, old and young voice of LSP parameter extractions of the voice, is carried out with the code book of N=256 sizes
Compression, the selected number of times of LSP parameters in record code book, and according to choosing number of times to be ranked up；By the LSP for choosing number of times most
Parameter coding is 00000000, and next is 00000001, is again 00000010.
3. the quantum compression method based on voice Multi-Band Excitation LSP parameters according to claim 1, its feature exists
In：The cryptoprinciple is that the number of times chosen is more, more comprising 0 number in parameter coding；Conversely, then comprising 1 number
It is more.
4. the quantum compression method based on voice Multi-Band Excitation LSP parameters according to claim 1, its feature exists
In：It is described right | ψ '>8 bits when measuring measurement result be 0, then Information State is | fghijkl0>, then proving will
The quantum bit of transmission is | fghijkl>It is typical state | ψlyp>, use | ψcomp1>Represent, channel sends | ψcomp1>；Measurement knot
Fruit is 1, then Information State is | zyxwvut1>, then prove that the quantum bit for sending is SARS kenel | zyxwvut>, i.e., | ψcomp2>,
Sent through quantum channel | ψcomp2>, wherein | ψcom>Meet
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410294051.XA CN104078047B (en) | 2014-06-21 | 2014-06-21 | Quantum compression method based on voice Multi-Band Excitation LSP parameters |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410294051.XA CN104078047B (en) | 2014-06-21 | 2014-06-21 | Quantum compression method based on voice Multi-Band Excitation LSP parameters |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104078047A CN104078047A (en) | 2014-10-01 |
| CN104078047B true CN104078047B (en) | 2017-06-06 |
Family
ID=51599269
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410294051.XA Active CN104078047B (en) | 2014-06-21 | 2014-06-21 | Quantum compression method based on voice Multi-Band Excitation LSP parameters |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104078047B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105959065B (en) * | 2016-06-28 | 2018-05-25 | 西安邮电大学 | The compression method and device of a kind of quantum information |
| CN112381233A (en) * | 2020-11-20 | 2021-02-19 | 北京百度网讯科技有限公司 | Data compression method and device, electronic equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0708435A1 (en) * | 1994-10-18 | 1996-04-24 | Matsushita Electric Industrial Co., Ltd. | Encoding and decoding apparatus of line spectrum pair parameters |
| EP1179820A2 (en) * | 2000-08-10 | 2002-02-13 | Mitsubishi Denki Kabushiki Kaisha | Method of coding LSP coefficients during speech inactivity |
| CN102495886A (en) * | 2011-11-28 | 2012-06-13 | 河南理工大学 | Method for searching fingerprint database based on quantum algorithm |
| CN103368658A (en) * | 2013-07-05 | 2013-10-23 | 西安邮电大学 | Four-photon entangled W state-based super-dense coding method for quantum signaling |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7315815B1 (en) * | 1999-09-22 | 2008-01-01 | Microsoft Corporation | LPC-harmonic vocoder with superframe structure |
| US7873513B2 (en) * | 2007-07-06 | 2011-01-18 | Mindspeed Technologies, Inc. | Speech transcoding in GSM networks |
-
2014
- 2014-06-21 CN CN201410294051.XA patent/CN104078047B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0708435A1 (en) * | 1994-10-18 | 1996-04-24 | Matsushita Electric Industrial Co., Ltd. | Encoding and decoding apparatus of line spectrum pair parameters |
| EP1179820A2 (en) * | 2000-08-10 | 2002-02-13 | Mitsubishi Denki Kabushiki Kaisha | Method of coding LSP coefficients during speech inactivity |
| CN102495886A (en) * | 2011-11-28 | 2012-06-13 | 河南理工大学 | Method for searching fingerprint database based on quantum algorithm |
| CN103368658A (en) * | 2013-07-05 | 2013-10-23 | 西安邮电大学 | Four-photon entangled W state-based super-dense coding method for quantum signaling |
Non-Patent Citations (5)
| Title |
|---|
| Quantum signaling transmission computing model and simulation analysis;Zhang Meiling et al.;《Automatic Control and Artificial Intelligence (ACAI 2012), International Conference on》;20120305;第1617-1620页 * |
| 一种低速率多模MBE线性预测语音编解码器;梁彦霞 等;《西安电子科技大学学报(自然科学版)》;20100630;第37卷(第3期);第417-422页 * |
| 一种应用于量子数据压缩的么正变换生成方法;梁彦霞 等;《物理学报》;20131020;第62卷(第20期);第200305-1至200305-5页 * |
| 线谱频率参数的快速、低存储矢量量化;梁彦霞 等;《吉林大学学报(工学版)》;20120131;第42卷(第1期);第223-227页 * |
| 量子语音多带激励算法;梁彦霞 等;《物理学报》;20140620;第63卷(第12期);第120301-1至120301-5页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104078047A (en) | 2014-10-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10341048B2 (en) | Channel encoding and decoding method and apparatus | |
| RU2522020C1 (en) | Hierarchical audio frequency encoding and decoding method and system, hierarchical frequency encoding and decoding method for transient signal | |
| DE60024123T2 (en) | LPC HARMONIOUS LANGUAGE CODIER WITH OVERRIDE FORMAT | |
| EP1514355B1 (en) | Method and system for multi-rate lattice vector quantization of a signal | |
| RU2007141934A (en) | ADAPTIVE GROUPING OF PARAMETERS FOR IMPROVED ENCODING EFFICIENCY | |
| CN107481295B (en) | Image compression system of convolutional neural network based on dynamic byte length distribution | |
| CN101510424A (en) | Method and system for encoding and synthesizing speech based on speech primitive | |
| CN104078047B (en) | Quantum compression method based on voice Multi-Band Excitation LSP parameters | |
| Pal et al. | An efficient codebook initialization approach for LBG algorithm | |
| US20200043199A1 (en) | 3d point cloud data encoding/decoding method and apparatus | |
| CN102982807B (en) | Method and system for multi-stage vector quantization of speech signal LPC coefficients | |
| CN102646417B (en) | Method for encoding and decoding enhancement layer | |
| KR102014384B1 (en) | Apparatus and method for discriminating vocoder type | |
| WO2023179800A1 (en) | Communication receiving method and apparatus thereof | |
| CN101617361A (en) | The method and apparatus that the speed of the speech business that is used to encode reduces | |
| CN101266796A (en) | A quantified coding method and device | |
| TWI376960B (en) | Codebook generating method for image compression | |
| CN101604524B (en) | Stereo coding method, stereo coding device, stereo decoding method and stereo decoding device | |
| CN113554719B (en) | Image encoding method, decoding method, storage medium and terminal equipment | |
| Chang et al. | Dither-like data hiding in multistage vector quantization of MELP and G. 729 speech coding | |
| Vimala et al. | Enhanced ambtc for image compression using block classification and interpolation | |
| CN110958417A (en) | Method for removing compression noise of video call video based on voice clue | |
| Guo et al. | SNR-Adaptive Multi-Layer Semantic Communication for Speech | |
| EP3089390B1 (en) | Decoding method and apparatus for system code | |
| JP2018146652A (en) | Coding/decoding method of voice |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| DD01 | Delivery of document by public notice |
Addressee: XI'AN University OF POSTS & TELECOMMUNICATIONS Person in charge of patents Document name: payment instructions |
|
| DD01 | Delivery of document by public notice | ||
| DD01 | Delivery of document by public notice |
Addressee: XI'AN University OF POSTS & TELECOMMUNICATIONS Person in charge of patents Document name: Notice of Termination of Patent Rights |
|
| DD01 | Delivery of document by public notice |