CN108417220B - Voice signal coding and decoding methods based on agent model Volterra modeling - Google Patents
Voice signal coding and decoding methods based on agent model Volterra modeling Download PDFInfo
- Publication number
- CN108417220B CN108417220B CN201810142277.6A CN201810142277A CN108417220B CN 108417220 B CN108417220 B CN 108417220B CN 201810142277 A CN201810142277 A CN 201810142277A CN 108417220 B CN108417220 B CN 108417220B
- Authority
- CN
- China
- Prior art keywords
- voice signal
- chaos
- formula
- undetermined coefficient
- model
- 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
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000009432 framing Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 29
- 239000003795 chemical substances by application Substances 0.000 claims description 20
- 230000003044 adaptive effect Effects 0.000 claims description 16
- 230000000739 chaotic effect Effects 0.000 claims description 7
- 230000001537 neural effect Effects 0.000 claims description 6
- 238000000714 time series forecasting Methods 0.000 claims description 5
- 238000010187 selection method Methods 0.000 claims description 4
- 238000001514 detection method Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 230000006698 induction Effects 0.000 claims description 3
- 230000005039 memory span Effects 0.000 claims description 3
- 239000000203 mixture Substances 0.000 claims description 2
- 238000007781 pre-processing Methods 0.000 claims 1
- 238000005316 response function Methods 0.000 claims 1
- 238000007796 conventional method Methods 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000002203 pretreatment Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
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
- G10L19/04—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 using predictive techniques
-
- 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
- G10L19/04—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 using predictive techniques
- G10L19/16—Vocoder architecture
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)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
A kind of voice signal coding and decoding methods based on agent model Volterra modeling construct prediction model by being pre-processed to the chaos voice signal of input, with Volterra modeling method, determine chaos voice signal prediction model and encode, decoding step forms.Existing artificial bee colony algorithm is improved since the present invention uses, preemphasis, adding window, framing pretreatment are carried out to the chaos voice signal of input, establish chaos voice signal prediction model, determine the parameter in chaos voice signal prediction model, complete the coding of chaos voice signal, according to the data having after encoding, conventionally it is decoded.The present invention utilizes the chaos feature of voice signal, rapidly and accurately realizes and is encoded, decoded to chaos voice signal, and with step, simple, easy to accomplish, high accuracy for examination, can be used for encoding chaos voice signal, decode.
Description
Technical field
The invention belongs to calculate and applied technical field, and in particular to Chaotic time series forecasting model.
Background technique
In recent years, reaching its maturity with hardware device and mechanics of communication, has change for the efficiency of transmission of voice
It is required that.Studies have found that voice signal time series is nonlinear, and show as apparent chaotic characteristic.Utilize chaos spy
Property building voice signal prediction model be considered as a kind of outstanding feasible method.Most of researcher's construction one non-thread
Property prediction model be all directly using Volterra modeling method:
And it is cumbersome to do phase space reconfiguration process.And it need to be built on the basis of voice signal chaotic characteristic using evolution algorithm
Voice signal Chaotic time series forecasting model is found.Existing evolution algorithm efficiency is lower, to particular problem without specific aim.It is existing
Somebody's work ant colony algorithm computational efficiency is low, solving precision is insufficient, existing observation peak stage Search equation are as follows:
The information of each iteration cannot be fully utilized.
Summary of the invention
Technical problem to be solved by the present invention lies in the above-mentioned prior art is overcome, provide a kind of step it is simple,
It is easy to accomplish, speed is fast, accuracy rate it is high based on agent model Volterra modeling voice signal coding and decoding methods.
Solving technical solution used by above-mentioned technical problem is to comprise the steps of:
(1) the chaos voice signal of input is pre-processed
In the chaos voice signal of input, the uniform frame of waveform is found as analysis frame, preemphasis is carried out, adding window, divides
Frame pretreatment.
Above-mentioned adding window pretreatment is carried out using following window function:
N is limited positive integer in formula.
(2) prediction model is constructed with Volterra modeling method
By the information of step (1) analysis frame, chaos voice signal prediction model is established by formula (2):
U (n-i τ) is the analysis frame signal of input in formula, and m is that the memory span of Chaotic time series forecasting model is limited
Positive integer, h1(i) and h2(i, j) is undetermined coefficient, and u (n-i τ) is the n-th-i τ sample of correspondence analysis frame, and n-i τ is step
(1) the sample serial number of analysis frame in, u (n-j τ) are the n-th-j τ sample of corresponding analysis frame, and n-j τ is analysis in step (1)
The sample serial number of frame, τ are delay times for limited positive integer, and j, n are limited positive integer.
(3) it determines chaos voice signal prediction model and encodes
The chaos voice signal of analysis frame in step (1) is determined into chaos voice signal institute with adaptive artificial bee colony algorithm
Corresponding delay time T, Embedded dimensions s, undetermined coefficient h1(i), undetermined coefficient h2(i, j), using agent model method as close
Like fitness function, Embedded dimensions s, delay time T, the undetermined coefficient h of high fitness are selected1(i) and undetermined coefficient h2(i,
J), it is obtained most as original fitness function by greedy selection method using the mean square error between predicted value and actual value
Good Embedded dimensions s, delay time T, undetermined coefficient h1(i) and h2(i, j) is substituted into above-mentioned formula (2), completes chaos voice signal
Coding.
(4) it decodes
By smallest embedding dimension number s, delay time τ, the undetermined coefficient h of the chaos voice signal of extraction1(i) and h2(i, j)
Substitution formula (2), obtains the prediction model to induction signal, according to the data having after encoding, is conventionally decoded.
Adaptive artificial bee colony algorithm in step of the present invention (3) are as follows:
ω is weight coefficient between (0,1) in formula, and c1, c2 are Studying factors 2,It is the random number of [- 1,1], xbest
For the global optimum nectar source of each iteration, xijFor current nectar source position, i is the serial number of nectar source vector, and j is respective components,
xneighborFor the neighbouring nectar source position in current nectar source, neighbor is that vector serial number in nectar source cannot be equal to i, and ω is by following two formula
Son determines:
ω=ωmin+ρ(ωmax-ωmin) (4)
ωminThe upper bound for being ω is 0.2, ωmaxThe lower bound for being ω is 0.9, a 2, and maxcyle is the largest the number of iterations
Being the largest the number of iterations for 2000, a 2, maxcyle is 1000 or 1500 or 2000.
Agent model method in step of the present invention (3) are as follows:
(1) the Embedded dimensions s in the chaos Phase Space Reconstruction of Speech Signals of analysis frame, delay time T are added to original
In Volterra model, the m in formula (1) is replaced with s.
(2) according to the model for introducing s and τ belonging to step (1), using the agent model method of adaptive artificial bee colony algorithm
Determine undetermined coefficient h1(i) and h2(i, j).
Use radial base neural net as approximate fitness function, by approximate fitness function and true fitness function
Models coupling uses, approximate fitness function are as follows:
K in formula (| | x-ci| |) it is used kernel function, aiFor the value to be assessed, ciFor radial base neural net
Central point, true fitness function are as follows:
Y in formulaiFor actual value,For predicted value, L is prediction length.
Determine best undetermined coefficient h1(i)、h2Whether (i, j), detection mean square error reach requirement, are not up to error requirements,
Iteration again.
Since the present invention is using existing artificial bee colony algorithm is improved, the chaos voice signal of input is carried out
Preemphasis, adding window, framing pretreatment, establish chaos voice signal prediction model, determine in chaos voice signal prediction model
Parameter, complete chaos voice signal coding, according to have coding after data, be conventionally decoded.This hair
The bright chaos feature using voice signal, rapidly and accurately realizes and is encoded, decoded to chaos voice signal, has step
Simply, easy to accomplish, high accuracy for examination can be used for encoding chaos voice signal, decode.
Detailed description of the invention
Fig. 1 is process flow chart of the invention.
Fig. 2 is the waveform diagram that embodiment 1 inputs chaos voice signal phonetic symbol [b].
Fig. 3 is the experimental result that embodiment 1 determines chaos voice signal prediction model and encodes.
Fig. 4 is the experimental result that embodiment 2 determines chaos voice signal prediction model and encodes.
Fig. 5 is the experimental result that embodiment 3 determines chaos voice signal prediction model and encodes.
Specific embodiment
The present invention is described in more detail with reference to the accompanying drawings and examples, but the present invention is not limited to following embodiment party
Formula.
Embodiment 1
By taking phonetic symbol [b] in the chaos voice signal chosen in standard pronunciation mark corpus as an example, it is based on agent model
The voice signal coding and decoding methods step (as shown in Figure 1) of Volterra modeling is as follows:
(1) the chaos voice signal of input is pre-processed
Fig. 2 is the waveform diagram of the chaos voice signal phonetic symbol [b] of input, in the chaos voice signal phonetic symbol [b] of input,
The uniform frame of waveform is found as analysis frame, carries out preemphasis, adding window, framing pretreatment, preemphasis is conventional method, using biography
Delivery function carries out preemphasis.
Above-mentioned adding window pretreatment is carried out using following window function:
N is limited positive integer in formula.
(2) prediction model is constructed with Volterra modeling method
The information of step (1) analysis frame is shown in Fig. 3, the present embodiment sample length that therefrom intercepted length is 400, by formula (2)
Establish chaos voice signal prediction model:
U (n-i τ) is the analysis frame signal of input in formula, and m is that the memory span of Chaotic time series forecasting model is limited
Positive integer, h1(i) and h2(i, j) is undetermined coefficient, and u (n-i τ) is the n-th-i τ sample of correspondence analysis frame, and n-i τ is step
(1) the sample serial number of analysis frame in, u (n-j τ) are the n-th-j τ sample of corresponding analysis frame, and n-j τ is analysis in step (1)
The sample serial number of frame, τ are delay times for limited positive integer, and j, n are limited positive integer.
(3) it determines chaos voice signal prediction model and encodes
The chaos voice signal of analysis frame in step (1) is determined into chaos voice signal institute with adaptive artificial bee colony algorithm
Corresponding delay time T, Embedded dimensions s, undetermined coefficient h1(i), undetermined coefficient h2(i, j), adaptive artificial bee colony algorithm are as follows:
ω is weight coefficient between (0,1) in formula, and c1, c2 are Studying factors 2,It is the random number of [- 1,1], xbest
For the global optimum nectar source of each iteration, xijFor current nectar source position, i is the serial number of nectar source vector, and j is respective components,
xneighborFor the neighbouring nectar source position in current nectar source, neighbor is that vector serial number in nectar source cannot be equal to i, and ω is by following two formula
Son determines:
ω=ωmin+ρ(ωmax-ωmin) (4)
ωminThe upper bound for being ω is 0.2, ωmaxThe lower bound for being ω is 0.9, a 2, and maxcyle is the largest the number of iterations
It is 2000.
Using agent model method as approximate fitness function, select the Embedded dimensions s of high fitness, delay time T,
Undetermined coefficient h1(i), undetermined coefficient h2(i, j), the agent model method of the present embodiment are as follows:
(1) the Embedded dimensions s in the chaos Phase Space Reconstruction of Speech Signals of analysis frame, delay time T are added to original
In Volterra model, the m in formula (1) is replaced with s.
(2) according to the model for introducing s and τ belonging to step (1), using the agent model method of adaptive artificial bee colony algorithm
Determine undetermined coefficient h1(i) and h2(i, j):
Use radial base neural net as approximate fitness function, by approximate fitness function and true fitness function
Models coupling uses, approximate fitness function are as follows:
K in formula (| | x-ci| |) it is used kernel function, aiFor the value to be assessed, ciFor radial base neural net
Central point, true fitness function are as follows:
Y in formulaiFor actual value,For predicted value, L is prediction length;
Determine best undetermined coefficient h1(i)、h2Whether (i, j), detection mean square error reach requirement, are not up to error requirements,
Iteration again.
Using the mean square error between predicted value and actual value as original fitness function, by greedy selection method,
Greedy selection method is conventional method, obtains smallest embedding dimension number s, delay time T, undetermined coefficient h1(i) and undetermined coefficient h2
(i, j) is substituted into above-mentioned formula (2), completes the coding of chaos voice signal.
With adaptive artificial bee colony algorithm obtain chaos voice signal phonetic symbol [b] corresponding to delay time T be 8, insertion dimension
Number s is 12, undetermined coefficient h in chaos voice signal prediction model1(i) and undetermined coefficient h2(i, j) is shown in Table 1, table 2, Fig. 3.
Best undetermined coefficient h in 1 embodiment 1 of table1(i)
h1(1) | h1(1) | h1(1) | h1(1) | h1(1) | h1(1) | h1(1) | h1(1) |
1 | ‐0.0020 | ‐0.0531 | ‐0.0898 | ‐0.1363 | 0.0555 | 0.6349 | ‐0.0617 |
Best undetermined coefficient h in 2 embodiment 1 of table2(i, j)
h2(i, j) | I=1 | I=2 | I=3 | I=4 | I=5 | I=6 | I=7 | I=8 |
J=1 | 0.8258 | -0.4758 | 0.2718 | 1 | -1 | 0.1292 | -1 | 0.7767 |
J=2 | 0.0449 | -0.0179 | 0.1362 | -0.1184 | 1 | 0.3567 | -0.3045 | |
J=3 | 0.5248 | 0.2685 | -0.9564 | 0.7436 | -0.3485 | 0.3652 | ||
J=4 | -0.9852 | 0.5326 | 0.2134 | 0.3452 | 0.2741 | |||
J=5 | 0.1245 | 0.5236 | -12354 | 1 | ||||
J=6 | -0.9654 | 0.1455 | 0.2542 | |||||
J=7 | 0.6532 | 0.8541 | ||||||
J=8 | 0.8745 |
By table 1, table 2, Fig. 3 as it can be seen that in chaos voice signal phonetic symbol [b] optimal embedding dimension s be 12, the optimum delay time
It is 8, undetermined coefficient h1(i)、h2When (i, j) is data in table, the worst error of sample cumulative is 0.199474, has reached error
Therefore range is output in the file of formulation.The file exported in figure is found, is substituted into above-mentioned formula (2), chaos voice is completed
The coding of signal.
(4) it decodes
By smallest embedding dimension number s, delay time τ, the undetermined coefficient h of phonetic symbol [b] in the chaos voice signal of extraction1(i)
And h2(i, j) substitute into formula (2), obtain the prediction model to induction signal, according to have coding after data, conventionally into
Row decoding.
Embodiment 2
By taking phonetic symbol [b] in the chaos voice signal chosen in standard pronunciation mark corpus as an example, it is based on agent model
Steps are as follows for the voice signal coding and decoding methods of Volterra modeling:
(1) the chaos voice signal of input is pre-processed
It is same as Example 1 that pre-treatment step is carried out to the chaos voice signal of input.
(2) prediction model is constructed with Volterra modeling method
It is same as Example 1 with Volterra modeling method building prediction model step.
(3) it determines chaos voice signal prediction model and encodes
The chaos voice signal of analysis frame in step (1) is determined into chaos voice signal institute with adaptive artificial bee colony algorithm
Corresponding delay time T, Embedded dimensions s, undetermined coefficient h1(i), undetermined coefficient h2(i, j), adaptive artificial bee colony algorithm are as follows:
ω is weight coefficient between (0,1) in formula, and c1, c2 are Studying factors 2,It is the random number of [- 1,1], xbest
For the global optimum nectar source of each iteration, xijFor current nectar source position, i is the serial number of nectar source vector, and j is respective components,
xneighborFor the neighbouring nectar source position in current nectar source, neighbor is that vector serial number in nectar source cannot be equal to i, and ω is by following two formula
Son determines:
ω=ωmin+ρ(ωmax-ωmin) (9)
ω in formulaminThe upper bound for being ω is 0.2, ωmaxThe lower bound for being ω is 0.9, a 2, and maxcyle is the largest iteration
Number is 1000.
Using agent model method as approximate fitness function, select the Embedded dimensions s of high fitness, delay time T,
Undetermined coefficient h1(i), undetermined coefficient h2The agent model method of (i, j), the present embodiment are same as Example 1.With adaptive artificial
Ant colony algorithm obtains that delay time T corresponding to chaos voice signal phonetic symbol [b] is 8, Embedded dimensions s is 12, chaos voice signal
Undetermined coefficient h in prediction model1(i) and undetermined coefficient h2(i, j) is shown in Table 3, table 4, Fig. 4.
Best undetermined coefficient h in 3 embodiment 2 of table1(i)
h1(1) | h1(1) | h1(1) | h1(1) | h1(1) | h1(1) | h1(1) | h1(1) |
1 | 1.1321 | 0.0672 | -0.4031 | 0.0203 | -0.2818 | 0.1010 | 0.2818 |
Best undetermined coefficient h in 4 embodiment 2 of table2(i, j)
Other steps are same as Example 1.
Complete the coding and decoding of chaos voice signal phonetic symbol [b].
Embodiment 3
By taking phonetic symbol [b] in the chaos voice signal chosen in standard pronunciation mark corpus as an example, it is based on agent model
Steps are as follows for the voice signal coding and decoding methods of Volterra modeling:
(1) the chaos voice signal of input is pre-processed
It is same as Example 1 that pre-treatment step is carried out to the chaos voice signal of input.
(2) prediction model is constructed with Volterra modeling method
It is same as Example 1 with Volterra modeling method building prediction model step.
(3) it determines chaos voice signal prediction model and encodes
The chaos voice signal of analysis frame in step (1) is determined into chaos voice signal institute with adaptive artificial bee colony algorithm
Corresponding delay time T, Embedded dimensions s, undetermined coefficient h1(i), undetermined coefficient h2(i, j), adaptive artificial bee colony algorithm are as follows:
ω is weight coefficient between (0,1) in formula, and c1, c2 are Studying factors 2,It is the random number of [- 1,1], xbest
For the global optimum nectar source of each iteration, xijFor current nectar source position, i is the serial number of nectar source vector, and j is respective components,
xneighborFor the neighbouring nectar source position in current nectar source, neighbor is that vector serial number in nectar source cannot be equal to i, and ω is by following two formula
Son determines:
ω=ωmin+ρ(ωmax-ωmim) (12)
ω in formulaminThe upper bound for being ω is 0.2, ωmaxThe lower bound for being ω is 0.9, a 2, and maxcyle is the largest iteration
Number is 1500.
Using agent model method as approximate fitness function, select the Embedded dimensions s of high fitness, delay time T,
Undetermined coefficient h1(i), undetermined coefficient h2The agent model method of (i, j), the present embodiment are same as Example 1.With adaptive artificial
Ant colony algorithm obtains that delay time T corresponding to chaos voice signal phonetic symbol [b] is 8, Embedded dimensions s is 12, chaos voice signal
Undetermined coefficient h in prediction model1(i) and undetermined coefficient h2(i, j) is shown in Table 5, table 6, Fig. 5.
Table 5 applies the best undetermined coefficient h in example 31(i)
h1(1) | h1(1) | h1(1) | h1(1) | h1(1) | h1(1) | h1(1) | h1(1) |
1 | 0.2119 | -0.4320 | -0.0315 | 0.0995 | 0.0014 | -0.1405 | 0.0898 |
Best undetermined coefficient h in 6 embodiment 3 of table2(i, j)
h2(i, j) | I=1 | I=2 | I=3 | I=4 | I=5 | I=6 | I=7 | I=8 |
J=1 | 0.2358 | -0.9652 | 0.2148 | 0.3541 | -1 | 0.7022 | -1 | 0.3354 |
J=2 | 0.6249 | -0.6931 | 0.3654 | -0.6944 | 0.2982 | 0.6367 | -0.4508 | |
J=3 | 0.9852 | 0.7564 | -0.2485 | 0.4267 | 0.5130 | 0.7452 | ||
J=4 | -0.3498 | 0.3215 | 0.3124 | 0.2347 | 0.7824 | |||
J=5 | 0.7545 | 0.1453 | -0.1154 | 1.2647 | ||||
J=6 | -0.5496 | 0.3265 | 0.3542 | |||||
J=7 | 0.3541 | 0.4516 | ||||||
J=8 | 0.1264 |
Other steps are same as Example 1.
Complete the coding and decoding of chaos voice signal phonetic symbol [b].
According to above-mentioned principle, different phonetic symbols in the chaos voice signal chosen in standard pronunciation mark corpus, using being based on
The voice signal coding and decoding methods of agent model Volterra modeling, can code and decode different phonetic symbols.
Claims (1)
1. a kind of voice signal coding and decoding methods based on agent model Volterra modeling, it is characterised in that by following step
Rapid composition:
(1) the chaos voice signal of input is pre-processed
In the chaos voice signal of input, the uniform frame of waveform is found as analysis frame, progress preemphasis, adding window, framing are pre-
Processing;
Above-mentioned adding window pretreatment is carried out using following window function:
N is limited positive integer in formula;
(2) prediction model is constructed with Volterra modeling method
By the information of step (1) analysis frame, chaos voice signal prediction model is established by formula (2):
In formula u (n-i τ) be input analysis frame signal, m be Chaotic time series forecasting model memory span be it is limited just
Integer, h1(i) and h2(i, j) is undetermined coefficient, and u (n-i τ) is the n-th-i τ sample of correspondence analysis frame, and n-i τ is step (1)
The sample serial number of middle analysis frame, u (n-j τ) are the n-th-j τ sample of corresponding analysis frame, and n-j τ is analysis frame in step (1)
Sample serial number, τ be delay time be limited positive integer, j,nFor limited positive integer;
(3) it determines chaos voice signal prediction model and encodes
Corresponding to the chaos voice signal of analysis frame in step (1) is determined chaos voice signal with adaptive artificial bee colony algorithm
Delay time T, Embedded dimensions s, undetermined coefficient h1(i), undetermined coefficient h2(i, j), using agent model method as approximate suitable
Response function selects Embedded dimensions s, delay time T, the undetermined coefficient h of high fitness1(i) and undetermined coefficient h2(i, j) is adopted
With the mean square error between predicted value and actual value as original fitness function, by greedy selection method, obtain best embedding
Enter dimension s, delay time T, undetermined coefficient h1 (i) and h2(i, j) is substituted into above-mentioned formula (2), completes the volume of chaos voice signal
Code;
Above-mentioned adaptive artificial bee colony algorithm are as follows:
ω is weight coefficient between (0,1) in formula, and c1, c2 are Studying factors 2,It is the random number of [- 1,1], xbestIt is every
The global optimum nectar source of secondary iteration, xijFor current nectar source position, i is the serial number of nectar source vector, and j is respective components, xneighbor
For the neighbouring nectar source position in current nectar source, neighbor is that vector serial number in nectar source cannot be equal to i, and ω is true by following two formula
It is fixed:
ω=ωmin+ρ(ωmax-ωmin) (4)
ωminThe upper bound for being ω is 0.2, ωmaxThe lower bound for being ω is 0.9, a 2, and maxcyle is the largest the number of iterations and is
It is 1000 or 1500 or 2000 that 2000, a 2, maxcyle, which are the largest the number of iterations,;
Above-mentioned agent model method are as follows:
(1) the Embedded dimensions s in the chaos Phase Space Reconstruction of Speech Signals of analysis frame, delay time T are added to original
In Volterra model, the m in formula (1) is replaced with s;
(2) it according to the model for introducing s and τ belonging to step (1), is determined using the agent model method of adaptive artificial bee colony algorithm
Undetermined coefficient h1(i) and h2(i, j):
Use radial base neural net as approximate fitness function, by approximate fitness function and true fitness function model
It is used in combination, approximate fitness function are as follows:
K in formula (| | x-ci| |) it is used kernel function, aiFor the value to be assessed, ciFor the center of radial base neural net
Point, true fitness function are as follows:
Y in formulaiFor actual value,For predicted value, L is prediction length;
Determine best undetermined coefficient h1(i)、h2Whether (i, j), detection mean square error reach requirement, are not up to error requirements, again
Iteration;
(4) it decodes
By smallest embedding dimension number s, delay time τ, the undetermined coefficient h of the chaos voice signal of extraction1(i) and h2(i, j) is substituted into
Formula (2), obtains the prediction model to induction signal, according to the data having after encoding, is conventionally decoded.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810142277.6A CN108417220B (en) | 2018-02-11 | 2018-02-11 | Voice signal coding and decoding methods based on agent model Volterra modeling |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810142277.6A CN108417220B (en) | 2018-02-11 | 2018-02-11 | Voice signal coding and decoding methods based on agent model Volterra modeling |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108417220A CN108417220A (en) | 2018-08-17 |
CN108417220B true CN108417220B (en) | 2019-06-25 |
Family
ID=63128600
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810142277.6A Active CN108417220B (en) | 2018-02-11 | 2018-02-11 | Voice signal coding and decoding methods based on agent model Volterra modeling |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108417220B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109745044B (en) * | 2018-12-19 | 2022-02-22 | 陕西师范大学 | Electroencephalogram signal coding and decoding method based on proxy model Volterra modeling |
CN109793514A (en) * | 2018-12-27 | 2019-05-24 | 陕西师范大学 | Mix the EEG signals coding and decoding methods for-Volterra model that leapfrogs |
CN110414031B (en) * | 2019-05-07 | 2021-10-22 | 深圳大学 | Method and device for predicting time sequence based on volterra series model, electronic equipment and computer readable storage medium |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2363557A (en) * | 2000-06-16 | 2001-12-19 | At & T Lab Cambridge Ltd | Method of extracting a signal from a contaminated signal |
US9716602B2 (en) * | 2013-07-08 | 2017-07-25 | Hughes Network Systems, Llc | System and method for iterative compensation for linear and nonlinear interference in system employing FTN symbol transmission rates |
CN104575512B (en) * | 2014-12-29 | 2018-06-26 | 南京航空航天大学 | Non-linear narrowband active noise controlling method based on Volterra wave filters |
CN105205838B (en) * | 2015-08-24 | 2018-02-27 | 重庆邮电大学 | A kind of vector quantization code book building method based on Chaos particle swarm optimization algorithm |
CN106782577A (en) * | 2016-11-11 | 2017-05-31 | 陕西师范大学 | A kind of voice signal coding and decoding methods based on Chaotic time series forecasting model |
-
2018
- 2018-02-11 CN CN201810142277.6A patent/CN108417220B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN108417220A (en) | 2018-08-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108417220B (en) | Voice signal coding and decoding methods based on agent model Volterra modeling | |
CN107844469A (en) | The text method for simplifying of word-based vector query model | |
CN109522411A (en) | A kind of writing householder method neural network based | |
CN108280112A (en) | Abstraction generating method, device and computer equipment | |
CN111931518A (en) | Translation model training method and device | |
CN107291836B (en) | Chinese text abstract obtaining method based on semantic relevancy model | |
CN111738020A (en) | Translation model training method and device | |
CN109977950A (en) | A kind of character recognition method based on mixing CNN-LSTM network | |
CN108388561B (en) | Neural network machine translation method and device | |
WO2019167600A1 (en) | Pseudo-bilingual data generation device, machine translation processing device, and pseudo-bilingual data generation method | |
CN106777073A (en) | The automatic method for correcting of wrong word and server in a kind of search engine | |
CN111126072B (en) | Method, device, medium and equipment for training Seq2Seq model | |
CN110942763A (en) | Voice recognition method and device | |
CN111401080A (en) | Neural machine translation method and neural machine translation device | |
CN111767697B (en) | Text processing method and device, computer equipment and storage medium | |
CN109525254A (en) | Convolutional code soft-decision decoding method based on deep learning | |
EP3525107A1 (en) | Conversational agent | |
CN106782577A (en) | A kind of voice signal coding and decoding methods based on Chaotic time series forecasting model | |
Zhou et al. | On generative-adversarial-network-based underwater acoustic noise modeling | |
CN109902313B (en) | Translation method and device, and translation model training method and device | |
Liu et al. | An improved fractal image coding based on the quadtree | |
CN110705279A (en) | Vocabulary selection method and device and computer readable storage medium | |
CN111797220B (en) | Dialog generation method, apparatus, computer device and storage medium | |
Zhang et al. | A RNN decoder for channel decoding under correlated noise | |
CN114330306A (en) | Deep learning-based password dictionary generation technology |
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 |