CN103310275B - Based on the Novel codebook design method of ant colony clustering and genetic algorithm - Google Patents

Abstract

Based on a Novel codebook design method for ant colony clustering and genetic algorithm, its step is as follows: step one: utilize the ant of different number to be distributed in randomly on the two-dimensional space of different size by training data, takes LF algorithm to carry out the generation of initial clustering; Step 2: the correction of the clustering cluster that initial clustering carries out according to code book size, ensures that the number of cluster is identical with the code book size of initial setting; Step 3: under the prerequisite that initial population successfully obtains, according to the basic procedure of genetic algorithm, carry out individual choice, intersection and mutation operation, until iteration stopping is met the optimum individual of requirement.The present invention not only overcomes the stronger shortcoming of the results relevance of initial selected and final design in LBG algorithm, avoids the inferior position that similar LBG algorithm is absorbed in local optimum simultaneously.It is applicable to the fields such as large quantity process, voice communication, pattern-recognition, IP phone.

Description

Based on the Novel codebook design method of ant colony clustering and genetic algorithm

Technical field

The invention provides a kind of Novel codebook design method based on ant colony clustering and genetic algorithm, what it related to is a kind of Novel codebook design method for the code book used of vector quantization in signal transacting, what be related specifically to is for the codebook design that line spectrum pair frequency parameter quantizes in MELP (Mixed Excitation Linear Prediction) encoding scheme, the method effectively can weaken the correlativity between design gained code book and training sequence, effectively prevent from being absorbed in local optimum in codebook design process simultaneously, be applicable at data dimension larger, the bit number that distribution is used for transmitting realizes data high precision when less quantizes.

Technical background

Along with the progressively development of society, the amount of communication data of mobile communication just progressively strengthens, traditional communication resource such as frequency spectrum cannot meet daily communication requirement, the concept of vector quantization is proposed for this kind of present studies personnel, vector quantization effectively can improve data-handling efficiency, reduce the communication resource that it takies, be thus widely used in the fields such as satellite communication, pattern-recognition, IP phone and large data compression.The concrete grammar of vector quantization is processed the data of a transmitting terminal multidimensional, by calculating the Euclidean distance (being referred to as code vector) between these data and the same dimension data of some, and then pick out between the two apart from minimum vector, and be used as substituting of pending data, what data were transmitted when transmitting is the order label of minor increment code vector in all code vectors, and data receiver carries out search index according to data label after receiving data, thus the optimum obtaining data handled by transmitting terminal substitutes.This shows that the design of code book is most important for the effect of vector quantization, effectively, accurate code book can ensure raw data as far as possible accurately to be recovered at data receiver, so the code book how obtaining effect optimum when limited training sequence just becomes a study hotspot in signal transacting field.

The algorithm of codebook design occur the earliest be LBG algorithm, proposed in 1980 by Linde, Buzo and Gray, this method does not need the probability distribution knowing training sequence, approaches optimum code book by taking certain iterative algorithm to training sequence.The key step of LBG algorithm is: 1, choose randomly from training sequence meet code book size number vector as initial code vector; 2, vectors all in training sequence is assigned in the first step around initial code vector according to the principle of minimum Eustachian distance, is referred to as clustering processing; 3, for the Vector operation centre of form of all training sequences that each cluster comprises, as new code vector; 4, calculate the distortion situation adopting all vectors of novel codebook quantification training sequence, if distortion reaches the thresholding of initial setting, stop, completing code book training, otherwise re-cover step 2 and 3.The advantage of LBG algorithm is that operation steps is simple, and the code book training time is shorter, most codebook design requirement can be met, but the limitation of this algorithm is obvious equally: the random selection method taked time first initialized causes the code book of final gained when initial selected is different, its quantification effect for training sequence difference is comparatively large, and namely the correlativity of optimum code book and initial code vector is stronger; Secondly, the design phase is easily absorbed in local optimum, causes and cannot obtain the final convergence effect needed.For these two shortcomings of LBG algorithm, many experts and scholars both domestic and external have carried out long research, and propose multiple effective codebook design method, effectively improve the effect of codebook design.

ZongboXie and JiuchaoFeng proposes a kind of codebook design method based on Mo Keer fuzzy clustering and gradient algorithm in paper " CodebookDesignForVectorQuantizationBasedonaKernelFuzzyLe arningAlgorithm ", in scheme, the Euclidean distance computing in LBG is taked the method for core distance, greatly reduce computation complexity, the effect of simulated effect show code the design promotes to some extent compared with LBG simultaneously.Genetic algorithm and simulated annealing are successfully applied in codebook design by H.B.Kekre and ChetanAgarwal in article " Codebookoptimizationusinggeneticalgorithmandsimulatedann ealing ", under the prerequisite additionally not increasing code book size, further optimization is carried out to code book, and the ability of searching optimum fully having used genetic algorithm good, avoid and be absorbed in local optimum, Be very effective.Chen Qian in paper " image vector based on hereditary LBG quantizes innovatory algorithm " by individual choice in genetic algorithmic procedures before all individualities carry out three iteration cluster operations, while making full use of LBG algorithm local search ability, improve the effect of codebook design and accelerate convergence of algorithm speed.Major part proposes now and the codebook design method used all and cannot weaken between initial selected and net result, correlativity to obtain and takes into account in raising design effect, although the algorithm had because be easy to is absorbed in the thresholding that code book that local optimum causes finally calculating can meet initial setting simultaneously, but the actual quantification effect that uses is poor, therefore propose one can effectively avoid being absorbed in local optimum, and the final code book codebook design method less with initial selected correlativity be the current problem being badly in need of researching and solving.

Summary of the invention

1, object:

In order to improve the quantification effect of code book, and then promote the degree of accuracy of vector quantization in actual signal process, traditional codebook design method has plenty of and solely utilizes ant colony clustering algorithm, make full use of its cluster result and the lower advantage of initial setting up correlativity, the code book obtained like this cannot meet application request in quantification effect, and cannot convergence state be reached sometimes within the limited training time, some schemes solely utilize genetic algorithm or simulated annealing, local optimum is absorbed in although avoid, but scheme still needs the selection of initial individuals in initial, between optimal result and initial selected, correlativity is close, the scheme also had sacrifices quantification effect for cost, reduce calculated amount in codebook design and time complexity, most method all cannot in quantification effect and real time complexity, obtain between initial relevance and take into account, the object of this invention is to provide a kind of Novel codebook design method based on ant colony clustering and genetic algorithm, it proposes a kind of codebook design schemes based on ant colony clustering and genetic algorithm, the program makes full use of ant colony clustering algorithm and the very low advantage of initial selected correlativity in the starting stage, in the two-dimensional space of different size, take the ant of different number to carry out the generation of initial clustering bunch, the centre of form calculating each clustering cluster is used as chromosome individual in the initial population in genetic manipulation, then take to select to it according to genetic algorithm, intersect and mutation operation, for the feature of real application data in practical operation, have employed dynamic interlace operation and mutation scheme, until the fitness of optimum individual and quantification effect reach the thresholding of setting, take full advantage of the outstanding ability of searching optimum of genetic algorithm avoiding while design process is absorbed in local optimum, effectively improve the effect of codebook design.

2, technical scheme

The present invention is characterized in: first the codebook design stage takes the LF algorithm in ant colony clustering algorithm to carry out the generation of initial clustering, the cardinal principle of this algorithm is the attractive force of copying occurring in nature ant to pile for the worker ant of carrying ant corpse is that the size of piling to ant is directly proportional, the larger attractive force of ant heap is larger, otherwise it is less, thus form a kind of positive feedback, be constantly change for the size of the two-dimensional space taked in this stage and the number of ant, then obtain different Clustering Effects; Then, the number of clustering cluster that may exist for the cluster obtained in the first step and the different shortcoming of code book size of setting, carry out the correction of clustering cluster; Finally, for the initial individuals obtained before, be used as the individuality of first generation population, take the basic procedure of genetic algorithm, perform the selection of genetic algorithm, intersection and mutation operation, the final shut-down operation when the fitness of optimum individual exceedes initial setting thresholding, exports the result being used as codebook design by optimum individual.

The Novel codebook design method that what Fig. 1 provided is based on ant colony clustering and genetic algorithm, a kind of Novel codebook design method based on ant colony clustering and genetic algorithm of the present invention, its concrete step is as follows:

Step one: utilize the ant of different number to be distributed in randomly on the two-dimensional space of different size by training data, takes LF algorithm to carry out the generation of initial clustering according to Fig. 2.

Step 2: the correction of the clustering cluster that the initial clustering carried out according to Fig. 3 carries out according to code book size, ensure that the number of cluster is identical with the code book size of initial setting.

Step 3: according to the basic procedure of Fig. 4 according to genetic algorithm under the prerequisite that initial population successfully obtains, carry out individual choice, intersection and mutation operation, until iteration stopping is met the optimum individual of requirement.

Wherein, that uses in step one takes LF algorithm to carry out the generation of initial clustering according to Fig. 2, and the parameter used in cluster process mainly comprises ant number m, two-dimensional space size Z, cluster radius s, distinctiveness ratio constant α, and constant value k _1,k ₂, carrying probability P _pput down probability P _d.

In step one, LF algorithm is taked to carry out the basic step of the generation of initial clustering as follows:

(1) setting initial ant number is m, and step-length is ma, and two-dimensional space size is initially Z, and step-length is za, and the data of training sequence be assigned to randomly on two-dimensional space under current ant number and two-dimensional space, all ants are all arranged to light condition.Now for the arbitrary element O in training sequence _i, its position X _iy _iall be in randomly in Z*Z space.

(2) for each ant, Cluster space is chosen a position randomly, if current ant is in light condition and current location has data exist, calculate the similarity of all data elements in current data element and the cluster radius centered by current location, stochastic generation one (0,1) numerical value in scope, if random number is greater than carrying probability, ant carries this data, and in current location, data empty.

(3) if current location does not have data and ant is in carrying state, the current ant of Similarity Measure that calculates in step (2) is used to put down probability for current data, calculating judges after completing: stochastic generation one (0,1) numerical value in scope, if random number exceedes put down probability, ant puts down data, and ant is arranged to light condition.

(4) position of ant is carried out fixed length randomly and is moved, and namely X or Y-coordinate plus-minus step-length, repeat mobile operation until the current position moved to is not occupied by other ant.

(5) repeat step (2)-(4) until to the equal complete operation of ant setting number, after complete operation, obtain the cluster result (code book size that the clusters number now obtained may reach with setting is different) for raw data.

(6) change initial parameter according to the ant number step-length ma set in step (1) and two-dimensional space step-length za, repeat step (2)-(5).

Wherein, the correction of the clustering cluster that the initial clustering carried out according to Fig. 3 in step 2 carries out according to code book size, to ensure that the number of clustering cluster and code book size are consistent.

The basic step of the correction of the clustering cluster that the initial clustering carried out in step 2 carries out according to code book size is as follows:

(1) Initial parameter sets: code book size M, the clusters number K obtained in Fig. 2.

(2) if M<K, the cluster data obtained in Fig. 2 is carried out descending sort according to the number of training sequence vector in each clustering cluster, then maximum clustering cluster and min cluster bunch are merged, replace original maximum clustering cluster, repeat union operation until clusters number is identical with the code book size of initial setting, i.e. M=K.

(3) if M>K, the cluster data obtained in Fig. 2 is carried out descending sort according to the number of training sequence vector in each clustering cluster, then the data in maximum clustering cluster are divided into two according to the size to this clustering cluster centre of form distance, generate two new clustering cluster, repeat fractured operation until clusters number is identical with the code book size of initial setting, i.e. M=K.

(4) centroid calculation is carried out for the current M an obtained clustering cluster, the centre of form is used as the code vector in initial codebook.

Wherein, the basic procedure of the genetic algorithm used in step 3, adopt genetic algorithm the initial population obtained before to be carried out selecting, intersect and the step of gene mutation operation, the parameter wherein used comprises Population Size N _{pop_size}, mutation probability P _m, crossover probability P _cwith iteration stopping thresholding ε.

Wherein, the basic procedure of the genetic algorithm adopted in step 3, its concrete step is as follows:

(1) respective fitness value is calculated for all individualities, then individuality is carried out descending sort according to fitness, judge whether the fitness of optimum individual is greater than iteration stopping thresholding, if be greater than iteration stopping thresholding, iteration work completes, enter step (4), on the contrary the step of entering (2).

(2) according to crossover probability P _cselect P before in ordering individual sequence _c* N _{pop_size}individuality directly enters the next generation, then the generation work that residue of future generation is individual is carried out: to ordering individual sequence, according to from above first with lastly to match, second principle of matching with penultimate above, two individualities of pairing are used as parent, then take single-point interlace operation to generate two new individualities, be placed in the next generation and be used as filial generation.Repeating crossover is until total individual number order meets Population Size.

(3) according to gene mutation probability P _mchoose P _m* N _{pop_size}individuality carries out mutation operation, and operating process is: each individuality is generated to the random number in (0, a 1) scope, if random number is greater than P _mthen current individual enters mutation operation, the gene on the chromosome of current individual is taked to the sudden change among a small circle, and sudden change numerical value is the function of individual fitness; If be less than P at random _mthen current individual is preserved, and does not enter mutation operation.Selection and mutation operation are all completed for all individualities, enters step (1).

(4) result of the maximum individuality of fitness as final codebook design is exported.

3, advantage and effect:

This Novel codebook design method based on ant colony clustering and genetic algorithm provided by the invention, not only overcomes the shortcoming that the results relevance of initial selected and final design in LBG algorithm is stronger, avoids the inferior position that similar LBG algorithm is absorbed in local optimum simultaneously.In codebook design process, first by using the LF algorithm in ant colony clustering algorithm, the ant of different number is used to carry out cluster to training data on different two-dimensional spaces, then feature that may be different from code book size for cluster gained clustering cluster number, adopt the amendment scheme introduced in Fig. 2, thus ensure that in the individuality obtained, the number of vector is identical with the size of code book, then the individuality obtained is used as the first generation initial in operatings of genetic algorithm individual, carry out the calculating of ideal adaptation degree, individual choice, intersect and gene mutation operation, the final fitness when optimum individual stops when being greater than iteration stopping thresholding, export the result of optimum individual now as codebook design, ensure that the carrying out of global optimization.

The present invention is applicable to the fields such as large quantity process, voice communication, pattern-recognition, IP phone, and the present invention mainly contains following advantage:

(1) LF algorithm is used to carry out the generation of initial clustering, the initial position of the advantage and last Clustering Effect and ant that take full advantage of LF algorithm is selected irrelevant, and the selection so just substantially overcoming different initial individuals in LBG algorithm affects for the strong correlation of final codebook design result.

(2) genetic algorithm is used to complete individual selection, intersection and mutation operation, qualified individuality is finally obtained by continuous iteration, take full advantage of the ability of searching optimum that genetic algorithm is stronger, overcome the shortcoming that LBG algorithm is easily absorbed in local optimum, ensure that final design codebooks obtains based on overall training sequence.

(3) there is not the problem being similar to LBG algorithm generation empty bag chamber in operation in genetic algorithm, and the great Bao chamber Split Method do not proposed for sky bag chamber, improves Clustering Effect in the design process.

Accompanying drawing explanation

Fig. 1 the method for the invention overview flow chart

Fig. 2 adopts LF algorithm to carry out the process flow diagram of the generation of initial clustering

The process flow diagram of the correction of the clustering cluster that Fig. 3 initial clustering carries out according to code book size

The basic flow sheet of Fig. 4 genetic algorithm

M: code book size

O _i: data element

Z*Z: the two-dimensional space of distribution

S: cluster radius

α: distinctiveness ratio constant

P _p: loading data probability

P _d: put down data probability

N _{pop_size}: Population Size

P _c: crossover probability

P _m: gene mutation probability

ε: iteration stopping thresholding

Embodiment

Below in conjunction with accompanying drawing 1,2,3,4, specifically introduce the Novel codebook design method given by the present invention for the line spectrum pair frequency parameter used in MELP (Mixed Excitation Linear Prediction) encoding scheme, wherein line spectrum pair frequency parameter is 10 dimensions.

A kind of Novel codebook design method based on ant colony clustering and genetic algorithm of the present invention, its concrete step is as follows:

Step one: default code book is of a size of 64, i.e. Population Size N _{pop_size}be 64.According to the scheme that Fig. 2 provides, the initial number first setting ant is 100, and step-length is 5, and two-dimensional space is of a size of 100, and step-length is 10, cluster radius s is 5, and distinctiveness ratio constant α is 2, constant k _1,k ₂be respectively 0.4,0.6, all ants are all arranged to light condition.The calculating of data similarity is carried out according to formula (1), and ant carries data probability and puts down the calculating of data probability according to formula (2), (3).Concrete formula is as follows:

$f\left(O_i\right)=\max \{0,\frac{1}{s^2}{\mathrm{\&Sigma;}}_{O_j,\mathrm{Neigh}(s*s)}[1-\frac{d(o_i,o_j)}{\mathrm{\&alpha;}}\left]\right\}...\left(1\right)$

$P_p\left(O_i\right)={\left(\frac{k_1}{k_1+f\left(o_i\right)}\right)}^2...\left(2\right)$

$P_d\left(O_i\right)=\left\{\begin{array}{c}2f\left(o_i\right),\mathrm{whenf}\left(o_i\right)<k_2\\ 1,f\left(o_i\right)\&GreaterEqual;k_2\end{array}\right....\left(3\right)$

Under the prerequisite that ant number and two-dimensional space size are determined, cluster process is:

(1) for i-th ant, random is distributed on two-dimensional space, if ant is in light condition and position has data exist, calculates similarity and carrying probability respectively according to formula (1), (2).Generate the random number in (0, a 1) scope, if random number is greater than carrying probability, ant carrying data, in current point, data empty, and ant is arranged to carrying state.If ant is in carrying state and current location does not have data, then calculate according to formula (3) and put down probability, generate one (0,1) random number in scope, if random number is greater than put down probability, ant puts down data, and store data in current point, ant becomes light condition.

(2) ant carries out position and moves, movement rule is X-axis not then X ± 1 to border, words then Y-coordinate ± 1 on the border that X-axis arrives, which kind of form added and subtracted is taked to depend on the generation of random number, number in [-1,1] scope for stochastic generation, if random number on the occasion of; get and add, random number be negative value then for subtracting, keep mobile until position is not occupied by other ant.

(3) repeat (1) (2) according to iterations, obtain the cluster result under current ant number and two-dimensional space size.

(4) according to step-size change ant number and the two-dimensional space size of setting, (1) (2) (3) are repeated, until obtain the cluster result of enough number of times.

Step 2: the correction initial clustering obtained in step one being carried out to clustering cluster according to code book size: if clustering cluster number is identical with code book size, the centre of form directly calculating each clustering cluster obtains code book; If clustering cluster number is greater than code book size, and how many clustering cluster is carried out descending sort according to the number of training sequence vector in affiliated bunch, then maximum clustering cluster and min cluster bunch are merged, replace original maximum clustering cluster, repeat union operation until clustering cluster number is identical with the code book size of initial setting; If clustering cluster number is less than code book size, the clustering cluster obtained carries out descending sort according to the number of training sequence vector in each clustering cluster, then the data in maximum clustering cluster are divided into two according to the size to this bunch of centre of form distance, generate two new clustering cluster, repeat fractured operation until clusters number is identical with the code book size of initial setting.Cluster result under all ant numbers and two-dimensional space size is all taken steps two operation.

Step 3: the basic procedure according to genetic algorithm after completing steps two operates, and has now obtained the individuality of the population first generation, uses the initial value of parameter: P in setting genetic algorithm _cbe 0.6, P _mbe 0.5, ε be 0.02.For each individuality, distortion summation when all vectors adopt this individuality to quantize in calculation training sequence, i.e. formula (4), ideal adaptation degree is the function of distortion summation, formula (5).

$\mathrm{dis}=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{(x_i-{\hat{x}}_i)}^2...\left(4\right)$

$\mathrm{fitness}\left(i\right)=\frac{1}{{\left({\mathrm{dis}}_i\right)}^2}...\left(5\right)$

(1) according to fitness, descending sort is carried out for all individualities, then carry out selecting and interlace operation, according to crossover probability P _cfront 38 individualities in ordering individual sequence are selected directly to enter the next generation, then the generation work that residue of future generation is individual is carried out: to ordering individual sequence, according to from above first with lastly to match, second principle of matching with penultimate above, two individualities of pairing are used as parent, then interlace operation is taked to generate two new individualities, the weighting interlace operation according to fitness is taked, namely according to formula (6) in crossover process:

$X_1=a*X+(1-a)*\hat{X}...\left(6\right)$

$X_2=(1-a)*X+a*\hat{X}$

Wherein the calculating of a weight is carried out according to formula (7).

$a=\frac{f_A^2}{{(f_A+f_{\underset{A}{^}})}^2}...\left(7\right)$

Newly-generated individuality is placed in the next generation and is used as filial generation.Repeating crossover is until total individual number order meets Population Size.

(2) gene mutation is taked to operate to all individualities, according to gene mutation probability, choose 32 individualities and carry out mutation operation, operating process is: each individuality is generated to the random number in (0, a 1) scope, if random number is greater than 0.5, current individual enters mutation operation, gene on the chromosome of current individual is taked to the sudden change among a small circle, five different random positions of sudden change position, sampling point place [1,10], sudden change numerical value is the random number in (-1,1) scope; If be less than 0.5 at random, then current individual is preserved, and does not enter mutation operation.Selection and mutation operation are all completed for all individualities.

(3) calculate the fitness of all individualities according to formula (4), (5), if the fitness of optimum individual is greater than repeatedly thresholding, then stop iteration, otherwise repetitive process (1) (2).

(4) net result of optimum individual as codebook design is exported.

In sum, a kind of Novel codebook design method based on ant colony clustering and genetic algorithm that the present invention proposes, innovation is 2 points: the LF algorithm that make use of ant colony clustering has on the one hand carried out the generation of initial clustering, make full use of the final Clustering Effect of LF and initial ant position and set advantage without Close relation, this just overcomes traditional LBG algorithm initial selected to the impact of net result; Genetic algorithm is utilized to carry out generation and the selection of optimum individual on the other hand, take full advantage of the ability of searching optimum that genetic algorithm is outstanding, overcome the shortcoming that LBG algorithm is very easily absorbed in local optimum simultaneously, can ensure that design codebooks is close to global optimum, gene mutation link have employed interlace operation based on fitness according to practical operation to line spectrum pair frequency parameter, change population development direction artificially, accelerate speed of convergence.Therefore, the present invention is on the basis of traditional ant colony clustering algorithm and genetic algorithm, fully take into account LF algorithm in practical operation in limiting time, possibly cannot arrive the shortcoming of convergence state, and take full advantage of the powerful global optimum's ability of genetic algorithm, be applicable to the codebook design process being applied to vector quantization early stage, the research for fields such as voice coding, data compression, pattern-recognitions has very large reference value.

Claims (2)

1., based on a Novel codebook design method for ant colony clustering and genetic algorithm, it is characterized in that: the step of the method is as follows:

Step one: utilize the ant of different number to be distributed in randomly on the two-dimensional space of different size by training data, takes LF algorithm to carry out the generation of initial clustering;

Step 2: the correction of the clustering cluster that initial clustering carries out according to code book size, ensures that the number of cluster is identical with the code book size of initial setting;

Step 3: under the prerequisite that initial population successfully obtains, according to the basic procedure of genetic algorithm, carry out individual choice, intersection and mutation operation, until iteration stopping is met the optimum individual of requirement;

Wherein, the correction of the clustering cluster that the initial clustering carried out in step 2 carries out according to code book size, to ensure that the number of clustering cluster and code book size are consistent;

The basic step of the correction of the clustering cluster that the initial clustering carried out in step 2 carries out according to code book size is as follows:

(1) Initial parameter sets: code book size M, clusters number K;

(2) if M<K, the cluster data obtained is carried out descending sort according to the number of training sequence vector in each clustering cluster, then maximum clustering cluster and min cluster bunch are merged, replace original maximum clustering cluster, repeat union operation until clusters number is identical with the code book size of initial setting, i.e. M=K;

(3) if M>K, the cluster data obtained is carried out descending sort according to the number of training sequence vector in each clustering cluster, then the data in maximum clustering cluster are divided into two according to the size to this clustering cluster centre of form distance, generate two new clustering cluster, repeat fractured operation until clusters number is identical with the code book size of initial setting, i.e. M=K;

(4) centroid calculation is carried out for the current M an obtained clustering cluster, the centre of form is used as the code vector in initial codebook;

Wherein, the basic procedure of the genetic algorithm used in step 3, adopt genetic algorithm the initial population obtained before to be carried out selecting, intersect and the step of gene mutation operation, the parameter wherein used comprises Population Size N _{pop_size}, mutation probability P _m, crossover probability P _cwith iteration stopping thresholding ε;

Wherein, the basic procedure of the genetic algorithm adopted in step 3, its concrete step is as follows:

(1) respective fitness value is calculated for all individualities, then individuality is carried out descending sort according to fitness, judge whether the fitness of optimum individual is greater than iteration stopping thresholding, if be greater than iteration stopping thresholding, iteration work completes, enter step (4), on the contrary the step of entering (2);

(2) according to crossover probability P _cselect P before in ordering individual sequence _c* N _{pop_size}individuality directly enters the next generation, then the generation work that residue of future generation is individual is carried out: to ordering individual sequence, according to from above first with lastly to match, second principle of matching with penultimate above, two individualities of pairing are used as parent, then single-point interlace operation is taked to generate two new individualities, be placed in the next generation and be used as filial generation, repeating crossover is until total individual number order meets Population Size;

(3) according to gene mutation probability P _mchoose P _m* N _{pop_size}individuality carries out mutation operation, and operating process is: each individuality is generated to the random number in (0, a 1) scope, if random number is greater than P _mthen current individual enters mutation operation, the gene on the chromosome of current individual is taked to the sudden change among a small circle, and sudden change numerical value is the function of individual fitness; If be less than P at random _mthen current individual is preserved, and does not enter mutation operation, all completes selection and mutation operation, enter step (1) for all individualities;

(4) result of the maximum individuality of fitness as final codebook design is exported.

2. a kind of Novel codebook design method based on ant colony clustering and genetic algorithm according to claim 1, it is characterized in that: the LF algorithm of taking used in step one carries out the generation of initial clustering, the parameter used in cluster process mainly comprises ant number m, two-dimensional space size Z, cluster radius s, distinctiveness ratio constant α, and constant value k ₁, k ₂, carrying probability P _pput down probability P _d;

In step one, LF algorithm is taked to carry out the basic step of the generation of initial clustering as follows:

(1) setting initial ant number is m, step-length is ma, two-dimensional space size is initially Z, step-length is za, under current ant number and two-dimensional space, the data of training sequence are assigned on two-dimensional space randomly, all ants are all arranged to light condition, now for the arbitrary element O in training sequence _i, its position X _iy _iall be in Z*Z space randomly;

(2) for each ant, Cluster space is chosen a position randomly, if current ant is in light condition and current location has data exist, calculate the similarity of all data elements in current data element and the cluster radius centered by current location, stochastic generation one (0,1) numerical value in scope, if random number is greater than carrying probability, ant carries this data, and in current location, data empty;

(3) if current location does not have data and ant is in carrying state, the current ant of Similarity Measure that calculates in step (2) is used to put down probability for current data, calculating judges after completing: stochastic generation one (0,1) numerical value in scope, if random number exceedes put down probability, ant puts down data, and ant is arranged to light condition;

(4) position of ant is carried out fixed length randomly and is moved, and namely X or Y-coordinate plus-minus step-length, repeat mobile operation until the current position moved to is not occupied by other ant;

(5) step (2)-(4) are repeated until to the equal complete operation of ant setting number, obtain the cluster result for raw data after complete operation, the clusters number now obtained may be different with setting the code book size reached;

(6) change initial parameter according to the ant number step-length ma set in step (1) and two-dimensional space step-length za, repeat step (2)-(5).