CN102088750B - Method and device for clustering propagation paths in multiple input multiple output (MIMO) technology - Google Patents

Abstract

The invention discloses a method and a device for clustering propagation paths in a multiple input multiple output (MIMO) technology. The method comprises the following steps: determining the number of candidate clusters and a corresponding clustering result of each candidate cluster; calculating a corresponding evaluation index of each clustering result; and determining an optimum clustering mode based on the corresponding evaluation index of each clustering result and a joint detection method, and taking the corresponding clustering result of the optimum clustering mode as the final clustering result. By utilizing the proposal provided by the invention, the propagation paths can be effectively clustered.

Description

Propagation footpath cluster-dividing method and device in a kind of Multiple Input Multiple Output

Technical field

The present invention relates to Multiple Input Multiple Output, particularly propagation footpath cluster-dividing method and the device in a kind of Multiple Input Multiple Output.

Background technology

In order to reach higher spectrum efficiency, at present, multiple-input, multiple-output (MIMO, Multiple InputMultiple Output) technology has become one of main flow candidate technologies of Next-Generation Wireless Communication Systems.Compare with traditional single-input single-output (SISO, Single Input Single Output) technology, Multiple Input Multiple Output takes full advantage of the degree of freedom of wireless channel on spatial domain, time-domain and frequency domain.

In the implementation procedure of existing Multiple Input Multiple Output, need in actual geographical environment, carry out the measurement of wireless channel, thereby obtain important wireless channel research foundation, such as channel impulse response etc., then further analyze and obtain large scale loss based on channel impulse response etc., sky, time, multipath fading parameter on frequency domain, and the statistical nature of large scale loss and multipath fading parameter etc.; Afterwards, further to statistical nature etc. rationally characterize, abstract and modeling, and set up the model of wireless channel based on this.

In actual electromagnetic propagation environment, due to scattering object, as the existence of trees and building etc., electromagnetic wave with distribute body and come in contact after, can there is certain change in its direction of propagation and amplitude.Conventionally characterize electromagnetic wave propagation with propagation footpath.Propagating footpath can characterize by a multi-Dimensional parameters collection, and this parameter set generally comprises: power, time delay, the angle of arrival and leave angle etc., each parameter characterization propagate the characteristic of footpath on different dimensions.The propagation footpath that adjacent scattering object or rough reflector produce has close parameter.Therefore, the propagation footpath with close parameter can be classified as to a set and study, this set is called bunch, and the process that different propagation footpaths is classified as bunch is called sub-clustering, and the assembly average of the parameter in all propagation footpath in each bunch is called a bunch heart.In above-mentioned reasonable sign, abstract and modeling process, can first sub-clustering be carried out in different propagation footpaths, then, based on each bunch, obtain respectively its bunch of intrinsic parameter and bunch between parameter, and bunch intrinsic parameter based on getting and bunch between parameter carry out subsequent treatment.Visible, how to carry out sub-clustering exactly, will directly have influence on objectivity and the accuracy of foundation of follow-up wireless model, but in prior art, also there is no a kind of generally acknowledged more effective cluster-dividing method.

Summary of the invention

In view of this, main purpose of the present invention is to provide the cluster-dividing method of the propagation footpath in a kind of Multiple Input Multiple Output, can realize for effective sub-clustering of propagating footpath.

Another object of the present invention is to provide the device of the propagation footpath sub-clustering in a kind of Multiple Input Multiple Output, can realize for effective sub-clustering of propagating footpath.

For achieving the above object, technical scheme of the present invention is achieved in that

A propagation footpath cluster-dividing method in Multiple Input Multiple Output, comprises the following steps:

Determine candidate's number of clusters, and determine sub-clustering result corresponding to each candidate's number of clusters;

Calculate every kind of evaluation index that sub-clustering result is corresponding;

According to every kind of evaluation index that sub-clustering result is corresponding, determine optimum sub-clustering mode based on joint detection method, using sub-clustering result corresponding optimum sub-clustering mode as final sub-clustering result.

Described definite sub-clustering result corresponding to each candidate's number of clusters comprises:

For each candidate's number of clusters K, carry out respectively following processing:

A, sorted by the ascending order of time delay in all propagation footpath for the treatment of sub-clustering, all propagation footpath for the treatment of sub-clustering after sequence is divided into K group in order;

B, the propagation footpath in every group with minimal time delay is defined as to the original cluster heart, and all original cluster hearts are divided into two set, one is effective bunch of heart set, and one is not inspection bunch heart set; Wherein, in described effective bunch of heart set, only comprise the original cluster heart of first grouping, the original cluster heart in effective bunch of heart set is called to the effective bunch of heart, will not check bunch heart in bunch heart set to be called the not inspection bunch heart;

C, do not check a bunch heart X for each in bunch heart set of inspection not, carry out respectively following processing:

The distance of each the effective bunch of heart in bunch heart X and described effective bunch of heart set is not checked in C1, calculating, and minimum range between the each distance calculating and predetermined bunch is compared, if the each distance calculating is all more than or equal to minimum range between described bunch, a described not inspection bunch heart X is included in described effective bunch of heart set, and end is for the processing of described not inspection bunch heart X, otherwise, execution step C2;

C2, choose described in the grouping of inspection bunch heart X place except the described not inspection bunch heart X propagation footpath of time delay minimum;

C3, calculate the distance of each the effective bunch of heart in selected propagation footpath and described effective bunch of heart set, and minimum range between the each distance calculating and described bunch is compared, if the each distance calculating is all more than or equal to minimum range between described bunch, selected propagation footpath is included in described effective bunch of heart set, and end is for the processing of described not inspection bunch heart X, otherwise, execution step C4;

The propagation footpath of time delay minimum in C4, the propagation footpath that is not selected in the grouping of inspection bunch heart X place described in choosing, and return to execution step C3;

D, will carry out the effective bunch of heart in effective bunch of heart set after treatment according to mode shown in step C as the initialization bunch heart;

E, according to the described initialization bunch heart, apart from dividing mode, all propagation footpath for the treatment of sub-clustering is divided into K bunch based on orthonormalization between footpath.

The method further comprises: if do not check all propagation footpath in bunch heart X place grouping all not meet the condition being included in described effective bunch of heart set described in step C, determine respectively every minimum value in the distance of propagating each the effective bunch of heart in footpath and described effective bunch of heart set, and select the maximum in the minimum value of determining, propagation footpath corresponding this maximum is included in described effective bunch of heart set.

Described step e comprises:

E1, calculate every and treat the propagation footpath of sub-clustering and the distance of each initialization bunch heart, by every propagation footpath for the treatment of sub-clustering be included into the minimum initialization bunch heart of distance corresponding bunch in, obtain K bunch;

The new bunch of heart of each bunch obtaining in E2, calculation procedure E1: ${\stackrel{\‾}{c}}_k=\frac{{\mathrm{\Σ}}_{l=1}^{L_k}(P_l\·X_l^k)}{{\mathrm{\Σ}}_{l=1}^{L_k}{P}_l},$ Wherein, X _l ^krepresent to propagate parameter set corresponding to footpath, P for l article in k bunch _lrepresent the l article of energy of propagating footpath, L _krepresent the propagation footpath number in k bunch, the value of described k is more than or equal to 1 and be less than or equal to K; Whether the new bunch of heart determining each bunch is all less than predefined threshold value with the distance of the initialization bunch heart of this bunch, if so, and end process, otherwise, execution step E3;

E3, calculate every and treat the propagation footpath of sub-clustering and the distance of each new bunch of heart, by every propagation footpath for the treatment of sub-clustering be included into the new bunch of minimum heart of distance corresponding bunch, obtain K bunch;

E4, the new new bunch of heart of each bunch of dividing of calculating: ${\stackrel{\‾}{c}}_k=\frac{{\mathrm{\Σ}}_{l=1}^{L_k}(P_l\·X_l^k)}{{\mathrm{\Σ}}_{l=1}^{L_k}{P}_l},$ Whether the distance of determining the new bunch of heart of this each bunch of calculating and the new bunch of heart of front this bunch once calculating is all less than predefined threshold value, if so, and end process, otherwise, return to execution step E3.

Between described bunch, definite method of minimum range is:

Any every two distances of propagating between footpaths in all propagation footpath of sub-clustering are treated in calculating, and find out ultimate range wherein;

Divided by K, obtain minimum range between described bunch by described ultimate range.

Every kind of evaluation index corresponding to sub-clustering result of described calculating comprises: calculate every kind of CH value and DB value that sub-clustering result is corresponding.

CH value corresponding to every kind of sub-clustering result of described calculating comprises:

Calculate $\stackrel{\‾}{c}=\frac{{\mathrm{\Σ}}_{l=1}^L(P_l\·X_l)}{{\mathrm{\Σ}}_{l=1}^L{P}_l},$ Wherein, X _lrepresent to treat all L articles of l articles of parameter sets corresponding to propagation footpath of propagating in footpath of sub-clustering, P _lrepresent the l article of energy of propagating footpath;

Calculate $\mathrm{tr}\left(B\right)=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{L}_k\·\mathrm{MD}{(c_k,\stackrel{\‾}{c})}^2,$ Wherein, L _krepresent the number in the propagation footpath in k bunch, MD (c _k, c) represent that the initialization bunch heart of k bunch arrives the distance of described c;

Calculate $\mathrm{tr}\left(W\right)=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\underset{j=1}{\overset{L_k}{\mathrm{\Σ}}}\mathrm{MD}{(X_j^k,c_k)}^2,$ Wherein, X _j ^krepresent that j article in k bunch is propagated parameter set corresponding to footpath, MD (X _j ^k, c _k) represent that j article in k bunch is propagated the distance of footpath to the initialization bunch heart of k bunch;

Calculate $\mathrm{CH}\left(K\right)=\frac{\mathrm{tr}\left(B\right)/(K-1)}{\mathrm{tr}\left(W\right)/(L-K)},$ Described CH (K) is every kind of CH value that sub-clustering result is corresponding.

DB value corresponding to every kind of sub-clustering result of described calculating comprises:

Calculate $S_k=\frac{1}{L_k}\underset{l=1}{\overset{L_k}{\mathrm{\Σ}}}\mathrm{MD}(X_l^k,c_k),$ Wherein, L _krepresent the number in the propagation footpath in k bunch, X _l ^krepresent that l article in k bunch is propagated parameter set corresponding to footpath, MD (X _l ^k, c _k) represent that l article in k bunch is propagated the distance of footpath to the initialization bunch heart of k bunch;

Calculate d _ij=MD (c _i, c _j), wherein, MD (c _i, c _j) represent the distance of the initialization bunch heart with the initialization bunch heart of individual bunch of j of i bunch; The value of described k, i and j is all more than or equal to 1 and be less than or equal to K;

Calculate $R_i=\underset{j=1,...,K,j\≠i}{\max }\left\{\frac{S_i+S_j}{d_{\mathrm{ij}}}\right\},$ Wherein, max represents to get maximum;

Calculate $\mathrm{DB}\left(K\right)=\frac{1}{K}\underset{i=1}{\overset{K}{\mathrm{\Σ}}}{R}_i,$ Described DB (K) is every kind of DB value that sub-clustering result is corresponding.

Describedly determine that based on joint detection method optimum sub-clustering mode comprises:

Determine the minimum value in DB value corresponding to every kind of sub-clustering result, this minimum value is multiplied by predefined proportionality constant t, obtain both products;

From DB value corresponding to every kind of sub-clustering result, select the DB value that is less than or equal to described product, candidate's number of clusters corresponding to each DB value of selecting is included in set F;

Determine the maximum in the CH value that each candidate's number of clusters of set in F is corresponding, candidate's number of clusters corresponding this maximum is defined as to optimum sub-clustering mode.

Every dimension parameter of all L bars for the treatment of sub-clustering being propagated to footpath forms respectively a parameter set, obtains altogether M parameter set, and described M represents to propagate the parameter dimension in footpath;

Calculate i parameter set x ⁱwith j parameter set x ^jbetween correlation $C_{\mathrm{ij}}=\frac{\underset{l=1}{\overset{L}{\mathrm{\Σ}}}(x_l^i-\stackrel{\‾}{x^i})(x_l^j-\stackrel{\‾}{x^j})}{\sqrt{\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{(x_l^i-\stackrel{\‾}{x^i})}^2}\sqrt{\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{(x_l^j-\stackrel{\‾}{x^j})}^2}},$ Wherein, x _l ⁱrepresent set x ⁱin l sample, x ⁱrepresent set x ⁱmean value, x _l ^jrepresent set x ^jin l sample, x ^jrepresent set x ^jmean value; The value of i and j is all more than or equal to 1 and be less than or equal to M;

Structure sample covariance matrix R, the element of the capable j row of i of this matrix is C _ij, obtain

The inverse matrix R of compute matrix R ^-1;

Calculate apart from d _{i ' j '}=(X _{i '}-X _{j '}) ' R ^-1(X _{i '}-X _{j '}), wherein, X _{i '}represent the parameter set of bunch heart of individual bunch of i ' article of propagation footpath or i ', symbol ' expression matrix transpose operation, X _{j '}represent the parameter set of bunch heart of individual bunch of j ' article of propagation footpath or j '; Work as X _{i '}, X _{j '}while being the parameter set of propagating footpath, described d _{i ' j '}represent to propagate the distance between footpath, work as X _{i '}, X _{j '}while being the parameter set of bunch heart, described d _{i ' j '}represent the distance between bunch heart, work as X _{i '}, X _{j '}in one for propagating the parameter set in footpath, when another is the parameter set of bunch heart, described d _{i ' j '}represent to propagate footpath to the distance between bunch heart.

A propagation footpath sub-clustering device in Multiple Input Multiple Output, comprising:

The first determination module, for determining candidate's number of clusters, and determines sub-clustering result corresponding to each candidate's number of clusters;

Computing module, for calculating every kind of evaluation index that sub-clustering result is corresponding;

The second determination module, for the evaluation index that sub-clustering result is corresponding according to every kind, determines optimum sub-clustering mode based on joint detection method, using sub-clustering result corresponding optimum sub-clustering mode as final sub-clustering result.

Described the first determination module comprises:

The first determining unit, for determining candidate's number of clusters;

The second determining unit, for for each candidate's number of clusters K, determines respectively its corresponding initialization bunch heart, comprising:

A, sorted by the ascending order of time delay in all propagation footpath for the treatment of sub-clustering, all propagation footpath for the treatment of sub-clustering after sequence is divided into K group in order;

B, the propagation footpath in every group with minimal time delay is defined as to the original cluster heart, and all original cluster hearts are divided into two set, one is effective bunch of heart set, and one is not inspection bunch heart set; Wherein, in described effective bunch of heart set, only comprise the original cluster heart of first grouping, the original cluster heart in effective bunch of heart set is called to the effective bunch of heart, will not check bunch heart in bunch heart set to be called the not inspection bunch heart;

C, do not check a bunch heart X for each in bunch heart set of inspection not, carry out respectively following processing:

The distance of each the effective bunch of heart in bunch heart X and described effective bunch of heart set is not checked in C1, calculating, and minimum range between the each distance calculating and predetermined bunch is compared, if the each distance calculating is all more than or equal to minimum range between described bunch, a described not inspection bunch heart X is included in described effective bunch of heart set, and end is for the processing of described not inspection bunch heart X, otherwise, execution step C2;

C2, choose described in the grouping of inspection bunch heart X place except the described not inspection bunch heart X propagation footpath of time delay minimum;

C3, calculate the distance of each the effective bunch of heart in selected propagation footpath and described effective bunch of heart set, and minimum range between the each distance calculating and described bunch is compared, if the each distance calculating is all more than or equal to minimum range between described bunch, selected propagation footpath is included in described effective bunch of heart set, and end is for the processing of described not inspection bunch heart X, otherwise, execution step C4;

The propagation footpath of time delay minimum in C4, the propagation footpath that is not selected in the grouping of inspection bunch heart X place described in choosing, and return to execution step C3;

D, will carry out the effective bunch of heart in effective bunch of heart set after treatment according to mode shown in step C as the initialization bunch heart;

The 3rd determining unit, for the initialization bunch heart corresponding according to each candidate's number of clusters K, apart from dividing mode, is divided into K bunch by all propagation footpath for the treatment of sub-clustering based on orthonormalization between footpath.

Described the second determining unit is further used for, if do not check all propagation footpath in bunch heart X place grouping all not meet the condition being included in described effective bunch of heart set described in step C, determine respectively every minimum value in the distance of propagating each the effective bunch of heart in footpath and described effective bunch of heart set, and select the maximum in the minimum value of determining, propagation footpath corresponding this maximum is included in described effective bunch of heart set.

Described the 3rd determining unit comprises:

The first computation subunit, treats the propagation footpath of sub-clustering and the distance of each initialization bunch heart for calculating every, by every propagation footpath for the treatment of sub-clustering be included into the minimum initialization bunch heart of distance corresponding bunch in, obtain K bunch;

The second computation subunit, for calculating the new bunch of heart of each bunch: ${\stackrel{\‾}{c}}_k=\frac{{\mathrm{\Σ}}_{l=1}^{L_k}(P_l\·X_l^k)}{{\mathrm{\Σ}}_{l=1}^{L_k}{P}_l},$ Wherein, X _l ^krepresent to propagate parameter set corresponding to footpath, P for l article in k bunch _lrepresent the l article of energy of propagating footpath, L _krepresent the propagation footpath number in k bunch; The value of described k is more than or equal to 1 and be less than or equal to K; Whether the new bunch of heart determining each bunch is all less than predefined threshold value with the distance of the initialization bunch heart of this bunch, if so, and end process, otherwise, notify the 3rd computation subunit to carry out self function;

Described the 3rd computation subunit, treats the propagation footpath of sub-clustering and the distance of each new bunch of heart for calculating every, by every propagation footpath for the treatment of sub-clustering be included into the new bunch of minimum heart of distance corresponding bunch, obtain K bunch, and notify the 4th computation subunit to carry out self function;

Described the 4th computation subunit, for calculating the new bunch of heart of each bunch of new division: ${\stackrel{\‾}{c}}_k=\frac{{\mathrm{\Σ}}_{l=1}^{L_k}(P_l\·X_l^k)}{{\mathrm{\Σ}}_{l=1}^{L_k}{P}_l},$ Whether the distance of determining the new bunch of heart of this each bunch of calculating and the new bunch of heart of front this bunch once calculating is all less than predefined threshold value, if so, and end process, otherwise, notify described the 3rd computation subunit to carry out self function.

Described the second determining unit is calculated the distances between any every two propagation footpaths for the treatment of in all propagation footpath of sub-clustering, and finds out ultimate range wherein, by described ultimate range divided by K, using phase division result as bunch between minimum range.

Described computing module comprises:

The first computing unit, for calculating every kind of CH value that sub-clustering result is corresponding, comprising:

Calculate $\stackrel{\‾}{c}=\frac{{\mathrm{\Σ}}_{l=1}^L(P_l\·X_l)}{{\mathrm{\Σ}}_{l=1}^L{P}_l},$ Wherein, X _lrepresent to treat all L articles of l articles of parameter sets corresponding to propagation footpath of propagating in footpath of sub-clustering, P _lrepresent the l article of energy of propagating footpath;

Calculate $\mathrm{tr}\left(B\right)=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{L}_k\·\mathrm{MD}{(c_k,\stackrel{\‾}{c})}^2,$ Wherein, L _krepresent the number in the propagation footpath in k bunch, MD (c _k, c) represent that the initialization bunch heart of k bunch arrives the distance of described c;

Calculate $\mathrm{tr}\left(W\right)=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\underset{j=1}{\overset{L_k}{\mathrm{\Σ}}}\mathrm{MD}{(X_j^k,c_k)}^2,$ Wherein, X _j ^krepresent that j article in k bunch is propagated parameter set corresponding to footpath, MD (X _j ^k, c _k) represent that j article in k bunch is propagated the distance of footpath to the initialization bunch heart of k bunch;

Calculate $\mathrm{CH}\left(K\right)=\frac{\mathrm{tr}\left(B\right)/(K-1)}{\mathrm{tr}\left(W\right)/(L-K)},$ Described CH (K) is every kind of CH value that sub-clustering result is corresponding;

The second computing unit, for calculating every kind of DB value that sub-clustering result is corresponding, comprising:

Calculate $S_k=\frac{1}{L_k}\underset{l=1}{\overset{L_k}{\mathrm{\Σ}}}\mathrm{MD}(X_l^k,c_k),$ Wherein, L _krepresent the number in the propagation footpath in k bunch, X _l ^krepresent that l article in k bunch is propagated parameter set corresponding to footpath, MD (X _l ^k, c _k) represent that l article in k bunch is propagated the distance of footpath to the initialization bunch heart of k bunch;

Calculate d _ij=MD (c _i, c _j), wherein, MD (c _i, c _j) represent the distance of the initialization bunch heart with the initialization bunch heart of individual bunch of j of i bunch; The value of described k, i and j is all more than or equal to 1 and be less than or equal to K;

Calculate $R_i=\underset{j=1,...,K,j\≠i}{\max }\left\{\frac{S_i+S_j}{d_{\mathrm{ij}}}\right\},$ Wherein, max represents to get maximum;

Calculate $\mathrm{DB}\left(K\right)=\frac{1}{K}\underset{i=1}{\overset{K}{\mathrm{\Σ}}}{R}_i,$ Described DB (K) is every kind of DB value that sub-clustering result is corresponding.

Described the second determination module comprises:

The 4th determining unit, for the minimum value of DB value corresponding to definite every kind of sub-clustering result, is multiplied by predefined proportionality constant t by this minimum value, obtains both products; From DB value corresponding to every kind of sub-clustering result, select the DB value that is less than or equal to described product, candidate's number of clusters corresponding to each DB value of selecting is included in set F;

The 5th determining unit, for the maximum in CH value corresponding to each candidate's number of clusters of definite set F, is defined as optimum sub-clustering mode by candidate's number of clusters corresponding this maximum, using sub-clustering result corresponding optimum sub-clustering mode as final sub-clustering result.

Visible, adopt technical scheme of the present invention, first determine candidate's number of clusters, and determine sub-clustering result corresponding to each candidate's number of clusters, then, calculate every kind of evaluation index that sub-clustering result is corresponding, finally, according to every kind of evaluation index that sub-clustering result is corresponding, determine optimum sub-clustering mode based on joint detection method, using sub-clustering result corresponding optimum sub-clustering mode as final sub-clustering result, thereby provide a kind of effective propagation footpath sub-clustering mode.

Brief description of the drawings

Fig. 1 is the flow chart of the inventive method embodiment.

Fig. 2 is the composition structural representation of apparatus of the present invention embodiment.

Embodiment

For problems of the prior art, the propagation footpath sub-clustering scheme in a kind of brand-new Multiple Input Multiple Output is proposed in the present invention, first determine candidate's number of clusters, and determine sub-clustering result corresponding to each candidate's number of clusters; Then, calculate every kind of evaluation index that sub-clustering result is corresponding; Finally, according to every kind of evaluation index that sub-clustering result is corresponding, determine optimum sub-clustering mode based on joint detection method, using sub-clustering result corresponding optimum sub-clustering mode as final sub-clustering result.

For making object of the present invention, technical scheme and advantage clearer, referring to the accompanying drawing embodiment that develops simultaneously, the present invention is described in further detail.

Fig. 1 is the flow chart of the inventive method embodiment.As shown in Figure 1, comprise the following steps:

Step 11: determine candidate's number of clusters.

Here candidate's number of clusters of mentioning, refers to all propagation footpath for the treatment of sub-clustering to be divided into how many bunches.The concrete value of candidate's number of clusters can rule of thumb be determined, or, also can adopt the simplest mode, suppose to treat that the propagation footpath number of sub-clustering is L, so, each integer that can be in the closed interval that 1 forms to L is all as candidate's number of clusters.The closed interval of supposing the candidate's number of clusters in the present embodiment is [K _min, K _max].

Step 12: determine the sub-clustering result that each candidate's number of clusters is corresponding.

In this step, for [K _min, K _max] in each candidate's number of clusters K, according to value order from small to large, respectively determine its corresponding sub-clustering result, specifically determine mode as follows:

A, sorted by the ascending order of time delay in all propagation footpath for the treatment of sub-clustering, all propagation footpath for the treatment of sub-clustering after sequence is divided into K group.

In the present embodiment, all propagation footpath for the treatment of sub-clustering after sequence can be equally divided into K group, if there is special circumstances, such as the number in all propagation footpath for the treatment of sub-clustering is 10, and the value of K is 3, be that both are aliquant, can arrange flexibly so and in every group after grouping, propagate footpath number, such as, first group comprises that 3 are propagated footpaths, the 2nd group comprises that 3 articles are propagated footpath, and the 3rd group comprises that 4 articles are propagated footpath; When aliquant, the propagation footpath number in first ensureing above each group is identical, and situation not enough or that have more is all put in last group.

B, the propagation footpath in every group with minimal time delay is defined as to original cluster heart c _k ⁰', k=1...K, and all original cluster hearts are divided into two set, one is effective bunch of heart set, one is not inspection bunch heart set; Wherein, in effective bunch of heart set, only comprise c ₁ ⁰', in inspection bunch heart set, do not comprise c ₀ ¹' ... c _k ⁰'; The original cluster heart in effective bunch of heart set is called to the effective bunch of heart, will does not check bunch heart in bunch heart set to be called the not inspection bunch heart.

C, for the not inspection bunch heart c of inspection in bunch heart set not ₂ ⁰', carry out following processing:

C1, calculating c ₂ ⁰' with effective bunch of heart set in the distance of each effective bunch of heart (initial condition only comprises c in effective bunch of heart set ₁ ⁰'), and minimum range between the each distance calculating and predetermined bunch is compared, if the each distance calculating be all more than or equal to bunch between minimum range, by c ₂ ⁰' be included in effective bunch of heart set, corresponding, by c ₂ ⁰in ' never inspection bunch heart set, delete, finish for c ₂ ⁰' processing, otherwise, execution step C2.

Wherein, between bunch, definite mode of minimum range is: calculate any every two distances of propagating between footpath in all propagation footpath for the treatment of sub-clustering, and find out ultimate range wherein; Afterwards, by this ultimate range divided by K, minimum range between the result obtaining is bunch.

C2, choose c ₂ ⁰in the grouping at ' place, remove c ₂ ⁰the propagation footpath of ' outer time delay minimum.

Because the propagation footpath in every group is all to arrange by the ascending order of time delay, the propagation footpath of therefore selecting in this step is next-door neighbour c ₂ ⁰' propagation footpath.

C3, calculate the distance of each the effective bunch of heart in selected propagation footpath and effective bunch of heart set, and by the each distance calculating and bunch between minimum range compare, minimum range between if the each distance calculating is all more than or equal to bunch, selected propagation footpath is included in effective bunch of heart set, and finishes for c ₂ ⁰' processing, otherwise, execution step C4.

C4, choose c ₂ ⁰the propagation footpath of time delay minimum in the propagation footpath not being selected in the grouping of ' place, and return to execution step C3.

Until find satisfied propagation footpath of being included into the effective bunch of condition in heart set.

But, if c ₂ ⁰all propagation footpath in the grouping of ' place does not all satisfy condition, can determine respectively every minimum value in the distance of propagating each the effective bunch of heart in footpath and effective bunch of heart set, and select the maximum in the minimum value of determining, propagation footpath corresponding this maximum is included in effective bunch of heart set.

For instance, suppose to comprise altogether in the effective bunch of heart set m the effective bunch of heart, and c ₂ ⁰the grouping at ' place comprises n bar propagation footpath, propagate footpath for every so, all can calculate m distance, find out this m the minimum value in distance, obtain altogether n minimum value, afterwards, then find out the maximum in this n minimum value, the corresponding propagation of this maximum footpath is included in effective bunch of heart set.

Afterwards, successively for c ₃ ⁰', c ₄ ⁰' ... c _k ⁰', perform step respectively C.

D, will carry out the effective bunch of heart in effective bunch of heart set after treatment according to mode shown in step C as initialization bunch heart c ₁ ⁰... c _k ⁰.

After the processing in step C, in effective bunch of heart set, will comprise K the effective bunch of heart, in the present embodiment, this K the effective bunch of heart is called to initialization bunch heart c ₁ ⁰... c _k ⁰, corresponding K grouping respectively.

The initialization bunch heart c that E, basis get ₁ ⁰... c _k ⁰, apart from dividing mode, all propagation footpath for the treatment of sub-clustering is divided into K bunch based on orthonormalization between footpath.

Between described footpath, orthonormalization can comprise apart from the specific implementation of dividing mode:

E1, calculate every and treat the propagation footpath of sub-clustering and the distance of each initialization bunch heart, by every propagation footpath for the treatment of sub-clustering be included into the minimum initialization bunch heart of distance corresponding bunch in, obtain K bunch;

The new bunch of heart of each bunch obtaining in E2, calculation procedure E1: ${\stackrel{\‾}{c}}_k=\frac{{\mathrm{\Σ}}_{l=1}^{L_k}(P_l\·X_l^k)}{{\mathrm{\Σ}}_{l=1}^{L_k}{P}_l},$ Wherein, X _l ^krepresent to propagate parameter set corresponding to footpath, P for l article in k bunch _lrepresent the l article of energy of propagating footpath, L _krepresent the propagation footpath number in k bunch; The value of k is more than or equal to 1 and be less than or equal to K; Whether the new bunch of heart determining each bunch is all less than predefined threshold value with the distance of the initialization bunch heart of this bunch, if so, and end process, otherwise, execution step E3; How each parameter is obtained is prior art;

E3, calculate every and treat the propagation footpath of sub-clustering and the distance of each new bunch of heart, by every propagation footpath for the treatment of sub-clustering be included into the new bunch of minimum heart of distance corresponding bunch, obtain K bunch;

E4, the new new bunch of heart of each bunch of dividing of calculating: ${\stackrel{\‾}{c}}_k=\frac{{\mathrm{\Σ}}_{l=1}^{L_k}(P_l\·X_l^k)}{{\mathrm{\Σ}}_{l=1}^{L_k}{P}_l},$ Whether the distance of determining the new bunch of heart of this each bunch of calculating and the new bunch of heart of front this bunch once calculating is all less than predefined threshold value, if so, and end process, otherwise, return to execution step E3.

Step 13: calculate every kind of evaluation index that sub-clustering result is corresponding.

Determine after the sub-clustering result that each candidate's number of clusters is corresponding, in this step, calculate every kind of CH value and DB value that sub-clustering result is corresponding, wherein, the account form of CH value is:

1) calculate $\stackrel{\‾}{c}=\frac{{\mathrm{\Σ}}_{l=1}^L(P_l\·X_l)}{{\mathrm{\Σ}}_{l=1}^L{P}_l},$ Wherein, X _lrepresent to treat all L articles of l articles of parameter sets corresponding to propagation footpath of propagating in footpath of sub-clustering, P _lrepresent the l article of energy of propagating footpath;

2) calculate $\mathrm{tr}\left(B\right)=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{L}_k\·\mathrm{MD}{(c_k,\stackrel{\‾}{c})}^2,$ Wherein, L _krepresent the number in the propagation footpath in k bunch, MD (c _k, c) represent that the initialization bunch heart of individual bunch of k is to the distance of c; C is commonly referred to the integral tufts heart;

3) calculate $\mathrm{tr}\left(W\right)=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\underset{j=1}{\overset{L_k}{\mathrm{\Σ}}}\mathrm{MD}{(X_j^k,c_k)}^2,$ Wherein, X _j ^krepresent that j article in k bunch is propagated parameter set corresponding to footpath, MD (X _j ^k, c _k) represent that j article in k bunch is propagated the distance of footpath to the initialization bunch heart of k bunch;

4) calculate $\mathrm{CH}\left(K\right)=\frac{\mathrm{tr}\left(B\right)/(K-1)}{\mathrm{tr}\left(W\right)/(L-K)},$ CH (K) is every kind of CH value that sub-clustering result is corresponding.The account form of DB value is:

1) calculate $S_k=\frac{1}{L_k}\underset{l=1}{\overset{L_k}{\mathrm{\Σ}}}\mathrm{MD}(X_l^k,c_k),$ Wherein, L _krepresent the number in the propagation footpath in k bunch, X _l ^krepresent that l article in k bunch is propagated parameter set corresponding to footpath, MD (X _l ^k, c _k) represent that l article in k bunch is propagated the distance of footpath to the initialization bunch heart of k bunch;

2) calculate d _ij=MD (c _i, c _j), wherein, MD (c _i, c _j) represent the distance of the initialization bunch heart with the initialization bunch heart of individual bunch of j of i bunch; The value of k, i and j is all more than or equal to 1 and be less than or equal to K;

3) calculate $R_i=\underset{j=1,...,K,j\≠i}{\max }\left\{\frac{S_i+S_j}{d_{\mathrm{ij}}}\right\},$ Wherein, max represents to get maximum;

4) calculate $\mathrm{DB}\left(K\right)=\frac{1}{K}\underset{i=1}{\overset{K}{\mathrm{\Σ}}}{R}_i,$ DB (K) is every kind of DB value that sub-clustering result is corresponding.

K in this step is identical with the implication of K in step 12, also represents candidate's number of clusters.

Step 14: according to every kind of evaluation index that sub-clustering result is corresponding, determine optimum sub-clustering mode based on joint detection method, using sub-clustering result corresponding optimum sub-clustering mode as final sub-clustering result.

The specific implementation of described joint detection method is:

1) determine the minimum value in DB value corresponding to every kind of sub-clustering result, this minimum value is multiplied by predefined proportionality constant t, obtain both products, the value of t can be decided according to the actual requirements; And select the DB value that is less than or equal to described product from DB value corresponding to every kind of sub-clustering result, candidate's number of clusters corresponding to each DB value of selecting is included in set F.

Be about to meet $\mathrm{DB}\left(K_i\right)\≤t\·\underset{K}{\min }\left\{\mathrm{DB}\left(K\right)\right\}$ Candidate's number of clusters be included into set F in, K _irepresent different candidate's number of clusters.

2) maximum in the CH value that the each candidate's number of clusters in definite set F is corresponding, is defined as optimum sub-clustering mode by candidate's number of clusters corresponding this maximum.

Calculate $K_{\mathrm{opt}}=\underset{K_i\∈F}{\arg \max }\left\{\mathrm{CH}\left(K_i\right)\right\},$ Wherein represent all CH (K _i) in K corresponding to maximum _ivalue, this K _ivalue must be arranged in set F.

Step 2) in the K that determines _ivalue is candidate's number of clusters corresponding to optimum sub-clustering mode, using the sub-clustering result of carrying out sub-clustering according to optimum sub-clustering mode as final sub-clustering result.

So far, completed flow process shown in the inventive method embodiment.

In above-described embodiment, the distance having related in a lot of situations between distance calculating and bunch heart and bunch heart between distance calculating, propagation footpath and bunch heart of propagating between footpath is calculated, and in the present embodiment, can adopt following account form:

1) every dimension parameter in all L bars propagation footpath for the treatment of sub-clustering is formed respectively to a set, obtain altogether M parameter set, M represents to propagate the parameter dimension in footpath;

2) calculate i parameter set x ⁱwith j parameter set x ^jbetween correlation $C_{\mathrm{ij}}=\frac{\underset{l=1}{\overset{L}{\mathrm{\Σ}}}(x_l^i-\stackrel{\‾}{x^i})(x_l^j-\stackrel{\‾}{x^j})}{\sqrt{\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{(x_l^i-\stackrel{\‾}{x^i})}^2}\sqrt{\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{(x_l^j-\stackrel{\‾}{x^j})}^2}},$ Wherein, x _l ⁱrepresent set x ⁱl sample, x ⁱrepresent set x ⁱmean value, x _l ^jrepresent set x ^jin l sample, x ^jrepresent set x ^jmean value; The value of i and j is all more than or equal to 1 and be less than or equal to M;

3) structure sample covariance matrix R, the element of the capable j row of i of this matrix is C _ij, obtain

4) the inverse matrix R of compute matrix R ^-1;

5) calculate apart from d _{i ' j '}=(X _{i '}-X _{j '}) ' R ^-1(X _{i '}-X _{j '}), wherein, X _{i '}represent the parameter set of bunch heart of individual bunch of i ' article of propagation footpath or i ', symbol ' expression matrix transpose operation, X _{j '}represent the parameter set of bunch heart of individual bunch of j ' article of propagation footpath or j '; Work as X _{i '}, X _{j '}while being the parameter set of propagating footpath, described d _{i ' j '}represent to propagate the distance between footpath, work as X _{i '}, X _{j '}while being the parameter set of bunch heart, described d _{i ' j '}represent the distance between bunch heart, work as X _{i '}, X _{j '}in one for propagating the parameter set in footpath, when another is the parameter set of bunch heart, d _{i ' j '}represent to propagate footpath to the distance between bunch heart.

Follow-up, final sub-clustering result that also can be based on determining in embodiment illustrated in fig. 1, the calculating of parameter and bunch intrinsic parameter between carrying out bunch, to provide basis for the foundation of follow-up wireless channel model.

Between bunch, the concrete account form of parameter and bunch intrinsic parameter is known in this field, such as,

In compute cluster, the time delay spreading parameter of time delay spreading parameter: k bunch is ${\mathrm{\σ}}_{\mathrm{\τ},k}=\sqrt{\underset{l=1}{\overset{L_k}{\mathrm{\Σ}}}{({\mathrm{\τ}}_l-{\mathrm{\μ}}_{\mathrm{\τ},k})}^2\·P_l/\underset{l=1}{\overset{L_k}{\mathrm{\Σ}}}{P}_l},$ Wherein, L _krepresent the number in the propagation footpath in k bunch, τ _lrepresent that l article in k bunch is propagated the time delay in footpath, P _lrepresent the power in this propagation footpath, μ _{τ, k}represent the time delay average of k bunch, ${\mathrm{\μ}}_{\mathrm{\τ},k}=\sqrt{\underset{l=1}{\overset{L_k}{\mathrm{\Σ}}}{\mathrm{\τ}}_l\·P_l/\underset{l=1}{\overset{L_k}{\mathrm{\Σ}}}{P}_l}.$

In compute cluster, the angle spread parameter of angle spreading parameter: k bunch is ${\mathrm{\&sigma;}}_{\mathrm{\&theta;},k}=\sqrt{\underset{l=1}{\overset{L_k}{\mathrm{\&Sigma;}}}{|{\mathrm{\&theta;}}_l-{\mathrm{\&mu;}}_{\mathrm{\&theta;},k})|}^2\&CenterDot;P_l/\underset{l=1}{\overset{L_k}{\mathrm{\&Sigma;}}}{P}_l},$ Wherein, σ _{θ, k}represent the angle spread of k bunch, θ _lrepresent that l article in k bunch is propagated the angle value in footpath, P _lrepresent the power in this propagation footpath, μ _{θ, k}represent the angle average of k bunch ${\mathrm{\&mu;}}_{\mathrm{\&theta;},k}=\sqrt{\underset{l=1}{\overset{L_k}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_l\&CenterDot;P_l/\underset{l=1}{\overset{L_k}{\mathrm{\&Sigma;}}}{P}_l},$ | θ _l-μ _{θ, k}| for: $|{\mathrm{\&theta;}}_l-{\mathrm{\&mu;}}_{\mathrm{\&theta;},k}|=\left\{\begin{array}{cc}2\mathrm{\&pi;}+({\mathrm{\&theta;}}_l-{\mathrm{\&mu;}}_{\mathrm{\&theta;},k}),& {\mathrm{\&theta;}}_l-{\mathrm{\&mu;}}_{\mathrm{\&theta;},k}<-\mathrm{\&pi;}\\ {\mathrm{\&theta;}}_l-{\mathrm{\&mu;}}_{\mathrm{\&theta;},k},& |{\mathrm{\&theta;}}_l-{\mathrm{\&mu;}}_{\mathrm{\&theta;},k}|\&le;\mathrm{\&pi;}\\ -2\mathrm{\&pi;}+({\mathrm{\&theta;}}_l-{\mathrm{\&mu;}}_{\mathrm{\&theta;},k}),& {\mathrm{\&theta;}}_l-{\mathrm{\&mu;}}_{\mathrm{\&theta;},k}>\mathrm{\&pi;}\end{array}\right.;$

In bunch, angle spreading parameter can further comprise that horizontal angle of arrival spreading parameter, level leave angle spreading parameter, vertical angle of arrival spreading parameter, and vertically leave angle and expand, concrete account form is identical, only need in computational process, bring different angle values into.

Based on said method, Fig. 2 is the composition structural representation of apparatus of the present invention embodiment.As shown in Figure 2, comprising:

The first determination module 21, for determining candidate's number of clusters, and determines sub-clustering result corresponding to each candidate's number of clusters;

Computing module 22, for calculating every kind of evaluation index that sub-clustering result is corresponding;

The second determination module 23, for the evaluation index that sub-clustering result is corresponding according to every kind, determines optimum sub-clustering mode based on joint detection method, using sub-clustering result corresponding optimum sub-clustering mode as final sub-clustering result.

Wherein, in the first determination module 21, can specifically comprise:

The first determining unit 211, for determining candidate's number of clusters;

The second determining unit 222, for for each candidate's number of clusters K, determines respectively its corresponding initialization bunch heart, comprising:

A, sorted by the ascending order of time delay in all propagation footpath for the treatment of sub-clustering, all propagation footpath for the treatment of sub-clustering after sequence is divided into K group in order;

B, the propagation footpath in every group with minimal time delay is defined as to the original cluster heart, and all original cluster hearts are divided into two set, one is effective bunch of heart set, and one is not inspection bunch heart set; Wherein, in effective bunch of heart set, only comprise the original cluster heart of first grouping, the original cluster heart in effective bunch of heart set is called to the effective bunch of heart, will not check bunch heart in bunch heart set to be called the not inspection bunch heart;

C, do not check a bunch heart X for each in bunch heart set of inspection not, carry out respectively following processing:

The distance of each the effective bunch of heart in bunch heart X and effective bunch of heart set is not checked in C1, calculating, and minimum range between the each distance calculating and predetermined bunch is compared, minimum range between if the each distance calculating is all more than or equal to bunch, will not check a bunch heart X to be included in effective bunch of heart set, and finish for the not processing of inspection bunch heart X, otherwise, execution step C2;

C2, choose not in the inspection bunch heart X place grouping propagation footpath of time delay minimum except inspection bunch heart X not;

C3, calculate the distance of each the effective bunch of heart in selected propagation footpath and effective bunch of heart set, and by the each distance calculating and bunch between minimum range compare, minimum range between if the each distance calculating is all more than or equal to bunch, selected propagation footpath is included in effective bunch of heart set, and finish for the not processing of inspection bunch heart X, otherwise, execution step C4;

C4, choose the propagation footpath of time delay minimum in the propagation footpath not being selected in inspection bunch heart X place grouping, and return to execution step C3;

D, will carry out the effective bunch of heart in effective bunch of heart set after treatment according to mode shown in step C as the initialization bunch heart;

The 3rd determining unit 213, for the initialization bunch heart corresponding according to each candidate's number of clusters K, apart from dividing mode, is divided into K bunch by all propagation footpath for the treatment of sub-clustering based on orthonormalization between footpath.

The second determining unit 212 can be further used for, if do not check all propagation footpath in bunch heart X place grouping all not meet the condition being included in described effective bunch of heart set in step C, determine respectively every minimum value in the distance of propagating each the effective bunch of heart in footpath and described effective bunch of heart set, and select the maximum in the minimum value of determining, propagation footpath corresponding this maximum is included in effective bunch of heart set.

In addition, in the 3rd determining unit 213, also can further comprise (for simplifying accompanying drawing, not shown):

The first computation subunit, treats the propagation footpath of sub-clustering and the distance of each initialization bunch heart for calculating every, by every propagation footpath for the treatment of sub-clustering be included into the minimum initialization bunch heart of distance corresponding bunch in, obtain K bunch;

The second computation subunit, for calculating the new bunch of heart of each bunch: ${\stackrel{\&OverBar;}{c}}_k=\frac{{\mathrm{\&Sigma;}}_{l=1}^{L_k}(P_l\&CenterDot;X_l^k)}{{\mathrm{\&Sigma;}}_{l=1}^{L_k}{P}_l},$ Wherein, X _l ^krepresent to propagate parameter set corresponding to footpath, P for l article in k bunch _lrepresent the l article of energy of propagating footpath, L _krepresent the propagation footpath number in k bunch; The value of k is more than or equal to 1 and be less than or equal to K; Whether the new bunch of heart determining each bunch is all less than predefined threshold value with the distance of the initialization bunch heart of this bunch, if so, and end process, otherwise, notify the 3rd computation subunit to carry out self function;

The 3rd computation subunit, treats the propagation footpath of sub-clustering and the distance of each new bunch of heart for calculating every, by every propagation footpath for the treatment of sub-clustering be included into the new bunch of minimum heart of distance corresponding bunch, obtain K bunch, and notify the 4th computation subunit to carry out self function;

The 4th computation subunit, for calculating the new bunch of heart of each bunch of new division: ${\stackrel{\&OverBar;}{c}}_k=\frac{{\mathrm{\&Sigma;}}_{l=1}^{L_k}(P_l\&CenterDot;X_l^k)}{{\mathrm{\&Sigma;}}_{l=1}^{L_k}{P}_l},$ Whether the distance of determining the new bunch of heart of this each bunch of calculating and the new bunch of heart of front this bunch once calculating is all less than predefined threshold value, if so, and end process, otherwise, notify the 3rd computation subunit to carry out self function.

Described the second determining unit is calculated the distances between any every two propagation footpaths for the treatment of in all propagation footpath of sub-clustering, and finds out ultimate range wherein, by this ultimate range divided by K, using phase division result as bunch between minimum range.

In computing module 22, can specifically comprise:

The first computing unit 221, for calculating every kind of CH value that sub-clustering result is corresponding, comprising:

Calculate $\stackrel{\&OverBar;}{c}=\frac{{\mathrm{\&Sigma;}}_{l=1}^L(P_l\&CenterDot;X_l)}{{\mathrm{\&Sigma;}}_{l=1}^L{P}_l},$ Wherein, X _lrepresent to treat all L articles of l articles of parameter sets corresponding to propagation footpath of propagating in footpath of sub-clustering, P _lrepresent the l article of energy of propagating footpath;

Calculate $\mathrm{tr}\left(B\right)=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{L}_k\&CenterDot;\mathrm{MD}{(c_k,\stackrel{\&OverBar;}{c})}^2,$ Wherein, L _krepresent the number in the propagation footpath in k bunch, MD (c _k, c) represent that the initialization bunch heart of individual bunch of k is to the distance of c;

Calculate $\mathrm{tr}\left(W\right)=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{L_k}{\mathrm{\&Sigma;}}}\mathrm{MD}{(X_j^k,c_k)}^2,$ Wherein, X _j ^krepresent that j article in k bunch is propagated parameter set corresponding to footpath, MD (X _j ^k, c _k) represent k bunch in j article propagate the distance of footpath to the initialization bunch heart of k bunch;

Calculate $\mathrm{CH}\left(K\right)=\frac{\mathrm{tr}\left(B\right)/(K-1)}{\mathrm{tr}\left(W\right)/(L-K)},$ CH (K) is every kind of CH value that sub-clustering result is corresponding;

The second computing unit 222, for calculating every kind of DB value that sub-clustering result is corresponding, comprising:

Calculate $S_k=\frac{1}{L_k}\underset{l=1}{\overset{L_k}{\mathrm{\&Sigma;}}}\mathrm{MD}(X_l^k,c_k),$ Wherein, L _krepresent the number in the propagation footpath in k bunch, X _l ^krepresent that l article in k bunch is propagated parameter set corresponding to footpath, MD (X _l ^k, c _k) represent that l article in k bunch is propagated the distance of footpath to the initialization bunch heart of k bunch;

Calculate d _ij=MD (c _i, c _j), wherein, MD (c _i, c _j) represent the distance of the initialization bunch heart with the initialization bunch heart of individual bunch of j of i bunch; The value of k, i and j is all more than or equal to 1 and be less than or equal to K;

Calculate $R_i=\underset{j=1,...,K,j\&NotEqual;i}{\max }\left\{\frac{S_i+S_j}{d_{\mathrm{ij}}}\right\},$ Wherein, max represents to get maximum;

Calculate $\mathrm{DB}\left(K\right)=\frac{1}{K}\underset{i=1}{\overset{K}{\mathrm{\&Sigma;}}}{R}_i,$ DB (K) is every kind of DB value that sub-clustering result is corresponding.

In the second determination module 23, can specifically comprise:

The 4th determining unit 231, for the minimum value of DB value corresponding to definite every kind of sub-clustering result, is multiplied by predefined proportionality constant t by this minimum value, obtains both products; From DB value corresponding to every kind of sub-clustering result, select the DB value that is less than or equal to described product, candidate's number of clusters corresponding to each DB value of selecting is included in set F;

The 5th determining unit 232, for the maximum in CH value corresponding to each candidate's number of clusters of definite set F, is defined as optimum sub-clustering mode by candidate's number of clusters corresponding this maximum, using sub-clustering result corresponding optimum sub-clustering mode as final sub-clustering result.

The specific works flow process of installing embodiment described in Fig. 2 please refer to the respective description in embodiment of the method shown in Fig. 1, repeats no more herein.

In sum, these are only preferred embodiment of the present invention, be not intended to limit protection scope of the present invention.Within the spirit and principles in the present invention all, any amendment of doing, be equal to replacement, improvement etc., within all should being included in protection scope of the present invention.

Claims (15)

1. the propagation footpath cluster-dividing method in Multiple Input Multiple Output, is characterized in that, comprises the following steps:

Determine candidate's number of clusters, and determine sub-clustering result corresponding to each candidate's number of clusters;

Calculate every kind of evaluation index that sub-clustering result is corresponding;

According to every kind of evaluation index that sub-clustering result is corresponding, determine optimum sub-clustering mode based on joint detection method, using sub-clustering result corresponding optimum sub-clustering mode as final sub-clustering result;

Wherein, described definite sub-clustering result corresponding to each candidate's number of clusters comprises:

For each candidate's number of clusters K, carry out respectively following processing:

A, sorted by the ascending order of time delay in all propagation footpath for the treatment of sub-clustering, all propagation footpath for the treatment of sub-clustering after sequence is divided into K group in order;

B, the propagation footpath in every group with minimal time delay is defined as to the original cluster heart, and all original cluster hearts are divided into two set, one is effective bunch of heart set, and one is not inspection bunch heart set; Wherein, in described effective bunch of heart set, only comprise the original cluster heart of first grouping, the original cluster heart in effective bunch of heart set is called to the effective bunch of heart, will not check bunch heart in bunch heart set to be called the not inspection bunch heart;

C, do not check a bunch heart X for each in bunch heart set of inspection not, carry out respectively following processing:

The distance of each the effective bunch of heart in bunch heart X and described effective bunch of heart set is not checked in C1, calculating, and minimum range between the each distance calculating and predetermined bunch is compared, if the each distance calculating is all more than or equal to minimum range between described bunch, a described not inspection bunch heart X is included in described effective bunch of heart set, and end is for the processing of described not inspection bunch heart X, otherwise, execution step C2;

C2, choose described in the grouping of inspection bunch heart X place except the described not inspection bunch heart X propagation footpath of time delay minimum;

C3, calculate the distance of each the effective bunch of heart in selected propagation footpath and described effective bunch of heart set, and minimum range between the each distance calculating and described bunch is compared, if the each distance calculating is all more than or equal to minimum range between described bunch, selected propagation footpath is included in described effective bunch of heart set, and end is for the processing of described not inspection bunch heart X, otherwise, execution step C4;

The propagation footpath of time delay minimum in C4, the propagation footpath that is not selected in the grouping of inspection bunch heart X place described in choosing, and return to execution step C3;

D, will carry out the effective bunch of heart in effective bunch of heart set after treatment according to mode shown in step C as the initialization bunch heart;

E, according to the described initialization bunch heart, apart from dividing mode, all propagation footpath for the treatment of sub-clustering is divided into K bunch based on orthonormalization between footpath.

2. method according to claim 1, it is characterized in that, the method further comprises: if do not check all propagation footpath in bunch heart X place grouping all not meet the condition being included in described effective bunch of heart set described in step C, determine respectively every minimum value in the distance of propagating each the effective bunch of heart in footpath and described effective bunch of heart set, and select the maximum in the minimum value of determining, propagation footpath corresponding this maximum is included in described effective bunch of heart set.

3. method according to claim 1, is characterized in that, described step e comprises:

E1, calculate every and treat the propagation footpath of sub-clustering and the distance of each initialization bunch heart, by every propagation footpath for the treatment of sub-clustering be included into the minimum initialization bunch heart of distance corresponding bunch in, obtain K bunch;

The new bunch of heart of each bunch obtaining in E2, calculation procedure E1: wherein, represent to propagate parameter set corresponding to footpath, P for l article in k bunch _lrepresent the l article of energy of propagating footpath, L _krepresent the propagation footpath number in k bunch, the value of described k is more than or equal to 1 and be less than or equal to K; Whether the new bunch of heart determining each bunch is all less than predefined threshold value with the distance of the initialization bunch heart of this bunch, if so, and end process, otherwise, execution step E3;

E3, calculate every and treat the propagation footpath of sub-clustering and the distance of each new bunch of heart, by every propagation footpath for the treatment of sub-clustering be included into the new bunch of minimum heart of distance corresponding bunch, obtain K bunch;

E4, the new new bunch of heart of each bunch of dividing of calculating: whether the distance of determining the new bunch of heart of this each bunch of calculating and the new bunch of heart of front this bunch once calculating is all less than predefined threshold value, if so, and end process, otherwise, return to execution step E3.

4. method according to claim 1, is characterized in that, between described bunch, definite method of minimum range is:

Any every two distances of propagating between footpaths in all propagation footpath of sub-clustering are treated in calculating, and find out ultimate range wherein;

Divided by K, obtain minimum range between described bunch by described ultimate range.

5. method according to claim 1, is characterized in that, every kind of evaluation index corresponding to sub-clustering result of described calculating comprises: calculate every kind of CH value and DB value that sub-clustering result is corresponding.

6. method according to claim 5, is characterized in that, CH value corresponding to every kind of sub-clustering result of described calculating comprises:

Calculate wherein, X _lrepresent to treat all L articles of l articles of parameter sets corresponding to propagation footpath of propagating in footpath of sub-clustering, P _lrepresent the l article of energy of propagating footpath;

Calculate wherein, L _krepresent the number in the propagation footpath in k bunch, described in representing that the initialization bunch heart of k bunch arrives distance;

Calculate wherein, represent that j article in k bunch is propagated parameter set corresponding to footpath, represent that j article in k bunch is propagated the distance of footpath to the initialization bunch heart of k bunch;

Calculate described CH(K) be every kind of CH value that sub-clustering result is corresponding.

7. method according to claim 5, is characterized in that, DB value corresponding to every kind of sub-clustering result of described calculating comprises:

Calculate wherein, L _krepresent the number in the propagation footpath in k bunch, represent that l article in k bunch is propagated parameter set corresponding to footpath, represent that l article in k bunch is propagated the distance of footpath to the initialization bunch heart of k bunch;

Calculate d _ij=MD (c _i, c _j), wherein, MD (c _i, c _j) represent the distance of the initialization bunch heart with the initialization bunch heart of individual bunch of j of i bunch; The value of described k, i and j is all more than or equal to 1 and be less than or equal to K;

Calculate wherein, max represents to get maximum;

Calculate described DB(K) be every kind of DB value that sub-clustering result is corresponding.

8. method according to claim 5, is characterized in that, describedly determines that based on joint detection method optimum sub-clustering mode comprises:

Determine the minimum value in DB value corresponding to every kind of sub-clustering result, this minimum value is multiplied by predefined proportionality constant t, obtain both products;

From DB value corresponding to every kind of sub-clustering result, select the DB value that is less than or equal to described product, candidate's number of clusters corresponding to each DB value of selecting is included in set F;

Determine the maximum in the CH value that each candidate's number of clusters of set in F is corresponding, candidate's number of clusters corresponding this maximum is defined as to optimum sub-clustering mode.

9. according to the method described in any one in claim 1～8, it is characterized in that,

Every dimension parameter of all L bars for the treatment of sub-clustering being propagated to footpath forms respectively a parameter set, obtains altogether M parameter set, and described M represents to propagate the parameter dimension in footpath;

Calculate i parameter set x ⁱwith j parameter set x ^jbetween correlation wherein, represent set x ⁱin l sample, represent set x ⁱmean value, represent set x ^jin l sample, represent set x ^jmean value; The value of i and j is all more than or equal to 1 and be less than or equal to M;

Structure sample covariance matrix R, the element of the capable j row of i of this matrix is C _ij, obtain

The inverse matrix R of compute matrix R ^-1;

Calculate apart from d _i'j'=(X _i'-X _j') ' R ^-1(X _i'-X _j'), wherein, X _i'represent the parameter set of bunch heart of individual bunch of i' article of propagation footpath or i', and symbol ' expression matrix transpose operation, X _j'represent the parameter set of bunch heart of individual bunch of j' article of propagation footpath or j'; Work as X _i', X _j'while being the parameter set of propagating footpath, described d _i'j'represent to propagate the distance between footpath, work as X _i', X _j'while being the parameter set of bunch heart, described d _i'j'represent the distance between bunch heart, work as X _i', X _j'in one for propagating the parameter set in footpath, when another is the parameter set of bunch heart, described d _i'j'represent to propagate footpath to the distance between bunch heart.

10. the propagation footpath sub-clustering device in Multiple Input Multiple Output, is characterized in that, comprising:

The first determination module, for determining candidate's number of clusters, and determines sub-clustering result corresponding to each candidate's number of clusters;

Computing module, for calculating every kind of evaluation index that sub-clustering result is corresponding;

The second determination module, for the evaluation index that sub-clustering result is corresponding according to every kind, determines optimum sub-clustering mode based on joint detection method, using sub-clustering result corresponding optimum sub-clustering mode as final sub-clustering result;

Wherein, described the first determination module comprises:

The first determining unit, for determining candidate's number of clusters;

The second determining unit, for for each candidate's number of clusters K, determines respectively its corresponding initialization bunch heart, comprising:

A, sorted by the ascending order of time delay in all propagation footpath for the treatment of sub-clustering, all propagation footpath for the treatment of sub-clustering after sequence is divided into K group in order;

B, the propagation footpath in every group with minimal time delay is defined as to the original cluster heart, and all original cluster hearts are divided into two set, one is effective bunch of heart set, and one is not inspection bunch heart set; Wherein, in described effective bunch of heart set, only comprise the original cluster heart of first grouping, the original cluster heart in effective bunch of heart set is called to the effective bunch of heart, will not check bunch heart in bunch heart set to be called the not inspection bunch heart;

C, do not check a bunch heart X for each in bunch heart set of inspection not, carry out respectively following processing:

The distance of each the effective bunch of heart in bunch heart X and described effective bunch of heart set is not checked in C1, calculating, and minimum range between the each distance calculating and predetermined bunch is compared, if the each distance calculating is all more than or equal to minimum range between described bunch, a described not inspection bunch heart X is included in described effective bunch of heart set, and end is for the processing of described not inspection bunch heart X, otherwise, execution step C2;

C2, choose described in the grouping of inspection bunch heart X place except the described not inspection bunch heart X propagation footpath of time delay minimum;

C3, calculate the distance of each the effective bunch of heart in selected propagation footpath and described effective bunch of heart set, and minimum range between the each distance calculating and described bunch is compared, if the each distance calculating is all more than or equal to minimum range between described bunch, selected propagation footpath is included in described effective bunch of heart set, and end is for the processing of described not inspection bunch heart X, otherwise, execution step C4;

The propagation footpath of time delay minimum in C4, the propagation footpath that is not selected in the grouping of inspection bunch heart X place described in choosing, and return to execution step C3;

D, will carry out the effective bunch of heart in effective bunch of heart set after treatment according to mode shown in step C as the initialization bunch heart;

The 3rd determining unit, for the initialization bunch heart corresponding according to each candidate's number of clusters K, apart from dividing mode, is divided into K bunch by all propagation footpath for the treatment of sub-clustering based on orthonormalization between footpath.

11. devices according to claim 10, it is characterized in that, described the second determining unit is further used for, if do not check all propagation footpath in bunch heart X place grouping all not meet the condition being included in described effective bunch of heart set described in step C, determine respectively every minimum value in the distance of propagating each the effective bunch of heart in footpath and described effective bunch of heart set, and select the maximum in the minimum value of determining, propagation footpath corresponding this maximum is included in described effective bunch of heart set.

12. according to the device described in claim 10 or 11, it is characterized in that, described the 3rd determining unit comprises:

The first computation subunit, treats the propagation footpath of sub-clustering and the distance of each initialization bunch heart for calculating every, by every propagation footpath for the treatment of sub-clustering be included into the minimum initialization bunch heart of distance corresponding bunch in, obtain K bunch;

The second computation subunit, for calculating the new bunch of heart of each bunch: wherein, represent to propagate parameter set corresponding to footpath, P for l article in k bunch _lrepresent the l article of energy of propagating footpath, L _krepresent the propagation footpath number in k bunch; The value of described k is more than or equal to 1 and be less than or equal to K; Whether the new bunch of heart determining each bunch is all less than predefined threshold value with the distance of the initialization bunch heart of this bunch, if so, and end process, otherwise, notify the 3rd computation subunit to carry out self function;

Described the 3rd computation subunit, treats the propagation footpath of sub-clustering and the distance of each new bunch of heart for calculating every, by every propagation footpath for the treatment of sub-clustering be included into the new bunch of minimum heart of distance corresponding bunch, obtain K bunch, and notify the 4th computation subunit to carry out self function;

Described the 4th computation subunit, for calculating the new bunch of heart of each bunch of new division: whether the distance of determining the new bunch of heart of this each bunch of calculating and the new bunch of heart of front this bunch once calculating is all less than predefined threshold value, if so, and end process, otherwise, notify described the 3rd computation subunit to carry out self function.

13. according to the device described in claim 10 or 11, it is characterized in that, described the second determining unit is calculated any every two distances of propagating between footpath in all propagation footpath for the treatment of sub-clustering, and find out ultimate range wherein, by described ultimate range divided by K, using phase division result as bunch between minimum range.

14. according to the device described in claim 10 or 11, it is characterized in that, described computing module comprises:

The first computing unit, for calculating every kind of CH value that sub-clustering result is corresponding, comprising:

Calculate wherein, X _lrepresent to treat all L articles of l articles of parameter sets corresponding to propagation footpath of propagating in footpath of sub-clustering, P _lrepresent the l article of energy of propagating footpath;

Calculate wherein, L _krepresent the number in the propagation footpath in k bunch, described in representing that the initialization bunch heart of k bunch arrives distance;

Calculate wherein, represent that j article in k bunch is propagated parameter set corresponding to footpath, represent that j article in k bunch is propagated the distance of footpath to the initialization bunch heart of k bunch;

Calculate described CH(K) be every kind of CH value that sub-clustering result is corresponding;

The second computing unit, for calculating every kind of DB value that sub-clustering result is corresponding, comprising:

Calculate wherein, L _krepresent the number in the propagation footpath in k bunch, represent that l article in k bunch is propagated parameter set corresponding to footpath, represent that l article in k bunch is propagated the distance of footpath to the initialization bunch heart of k bunch;

Calculate d _ij=MD (c _i, c _j), wherein, MD (c _i, c _j) represent the distance of the initialization bunch heart with the initialization bunch heart of individual bunch of j of i bunch; The value of described k, i and j is all more than or equal to 1 and be less than or equal to K;

Calculate wherein, max represents to get maximum;

Calculate described DB(K) be every kind of DB value that sub-clustering result is corresponding.

15. devices according to claim 14, is characterized in that, described the second determination module comprises:

The 4th determining unit, for the minimum value of DB value corresponding to definite every kind of sub-clustering result, is multiplied by predefined proportionality constant t by this minimum value, obtains both products; From DB value corresponding to every kind of sub-clustering result, select the DB value that is less than or equal to described product, candidate's number of clusters corresponding to each DB value of selecting is included in set F;

The 5th determining unit, for the maximum in CH value corresponding to each candidate's number of clusters of definite set F, is defined as optimum sub-clustering mode by candidate's number of clusters corresponding this maximum, using sub-clustering result corresponding optimum sub-clustering mode as final sub-clustering result.