WO2012116527A1

WO2012116527A1 - Segmented frequency domain method and device for realizing multi-path search

Info

Publication number: WO2012116527A1
Application number: PCT/CN2011/075775
Authority: WO
Inventors: 沈承科
Original assignee: 中兴通讯股份有限公司
Priority date: 2011-03-02
Filing date: 2011-06-15
Publication date: 2012-09-07
Also published as: CN102655420B; CN102655420A

Abstract

A segmented frequency domain method and device for realizing multi-path search are provided by the present invention. The present invention employs the frequency domain segmented search on every sub-segment by dividing a whole convolution segment into several shorter convolution segments, and further reduces the total computing complication. And if an FFT hardware accelerator is used to complete the search, since the segment processing is employed, a small size FFT can be used to replace a large size FFT to complete the search which needs to be completed only by the large size FFT, thus the hardware implementing complication is reduced, therefore the cost is reduced, and the original small size FFT module can be used repeatedly.

Description

Segmentation frequency domain method and device for realizing multipath search

The present invention relates to multipath search techniques, and more particularly to a segmentation frequency domain method and apparatus for implementing multipath search. Background technique

Code Division Multiple Access (CDMA) Multipath search is used to search for multipath delay and attenuation of the channel, which can be done by detecting the channel's impulse response (CIR). Multipath search for CDMA is usually done by time domain convolution. The time domain convolution can also be equivalently completed by the frequency domain product. Using the frequency domain method instead of the time domain convolution can greatly simplify the computational complexity of multipath search. The computational complexity of the frequency domain method and the fast Fourier used. The size of the leaf transform (FFT) is related.

With the increasing use of multi-mode soft-baseband platforms, chip-level processing of CDMA systems (or WCDMA, CDMA2000, etc.) is becoming more and more feasible with a general-purpose digital signal processing (DSP) or parallel array processor as the core. For CDMA multipath searchers, the computational complexity of frequency domain multiplication searches is much lower than the complexity of time domain direct convolutional search.

The CDMA multipath search time domain method generally comprises: convolving with a locally generated pseudo code and a baseband signal received by an antenna, and the result of the convolution is the impact response of the channel. The problem with the time domain convolution method is that the amount of computation is huge. 1 is a schematic diagram of a conventional frequency domain CDMA multipath search. As shown in FIG. 1, X is a CDMA antenna baseband signal chip sampling signal having a length of Lx chips; and a local pseudocode sequence length Lc. Assuming that the channel search length is L (unit: chip), then Lx = L + Lc.

The frequency domain method is to extend the X-padded length of Lx to _Lfft , _Lfft is the smallest integer greater than Lx, and satisfies the exponent of 2 (can be processed by FFT); Then, it is completed by the FFT operation with the scale of L _fft point. Trail search. The mathematical principle can be given by the formula (1): y = - IFFT {eii g\ FFT {G|l,*|} | x FFT {\} } (1)

In equation (1), χ is an input signal sequence after extending X-padding of length Lx to L _fft , FFT() is an FFT transform of length L _fft ; G[l , *] is a length of Lc The local pseudo-code sequence zero-padded to the pseudo-code sequence after L _fft ; IFFT0 is an inverse FFT of length L _fft , and the result of the inverse FFT is a vector y of length L _fft , the vector y is composed of two sub-vectors y0 And yl cascaded, wherein y0 is L, which is the impulse response of the channel. The block diagram of the existing frequency domain searcher is shown in Figure 2.

The frequency domain method greatly reduces the computational complexity. For example, if £ is Lc=256, L=128, then the time domain convolution requires 256*128=32768 complex multiplication; when using the frequency domain method, assuming L _fft = 512, only 512*log ₂ 512 is needed. +512=5120 complex multiplication. Summary of the invention

In view of this, the main object of the present invention is to provide a segmentation frequency domain method and apparatus for implementing CDMA multipath search based on the frequency domain method, which can further reduce the total computational complexity and reduce the cost.

In order to achieve the above object, the technical solution of the present invention is achieved as follows:

A segmentation frequency domain method for implementing multipath search, including:

Cutting the pseudo code sequence into a preset number of subsequence segments;

After the input signal sequence is padded with zeros, it is divided into a preset number of subsequence segments that are partially overlapped with each other;

Frequency domain search is performed separately for each segment, and the frequency domain search results of each segment are superimposed and merged.

The cutting the pseudo code sequence into a preset number of subsequence segments includes:

The pseudo code sequence c of length Lc is cut into a preset number of K subsequence segments, respectively c(l), c(2), ..., c(K), each segment is N, and satisfies K*N=Lc; N=L+LQ, where L. Is to make Let N = L + Lo be the smallest integer of the exponent of 2;

Where L is the time domain length of the searched channel, Lc is the length of the CDMA local pseudocode and its length is a power of 2, and L < Lc.

After the end of the input signal sequence is expanded by zero, the sub-sequence segment divided into a preset number of mutually overlapping portions includes:

The length of the input signal sequence X is (L+Lc), and the end of the input signal sequence X is zero-padded to a length that satisfies K*N+L+L. = (K+1)*N;

The input signal sequence after the zero-pad expansion is segmented into a preset number of K sub-sequences that are partially overlapped, respectively, x(l), x(2), ..., x(K), and each subsequence The length of the segment is 2N.

The frequency domain search for each segment is: using an existing non-segmented frequency domain search method.

The superimposing and combining the frequency domain search results of each segment includes: a frequency domain search result y outputting a preset number of K segments respectively. (l) , y ₀ (2), y. (K) All are superimposed and combined according to formula (2) to obtain the final frequency domain search result y ₀ .

A segmentation frequency domain device for implementing multipath search, comprising at least: a pseudo code sequence segmentation module, an input signal sequence segmentation module, a preset number of K frequency domain search modules, and a merge processing module, wherein

The pseudo code sequence segmentation module is configured to cut the pseudo code sequence into a preset number of subsequence segments; the input signal sequence segmentation module is configured to expand the input signal sequence end to zero, and then divide into a preset number of mutually overlapping portions. Subsequence segment;

Each frequency domain search module is configured to perform frequency domain search on the kth segment according to the k pseudocode subsequence segment from the pseudo code sequence segmentation module and the kth input signal subsequence segment from the input signal sequence segmentation module, And outputting the k-th segment search result to the merge processing module; wherein k is 1 or 2, ... or K;

The merging processing module is configured to superimpose and merge the frequency domain search results of each segment to obtain a final frequency domain search result. It can be seen from the above technical solution provided by the present invention that the present invention is used to split the entire convolution segment into a plurality of shorter convolution segments, and frequency domain segmentation search is performed on each sub segment, further reducing the total The computational complexity; Moreover, if done with an FFT hardware accelerator, a small-sized FFT can be used instead of the original search that requires a large-size FFT, which reduces the complexity of the hardware implementation, thereby reducing the cost and reusing the original Small size FFT module. DRAWINGS

1 is a schematic diagram of a conventional CDMA multipath search;

2 is a block diagram of a composition of a conventional frequency domain searcher;

3 is a flowchart of a method for implementing multipath search according to the present invention;

4 is a schematic diagram of implementation of pseudo-code data block segmentation in the multipath search method of the present invention; FIG. 5 is a schematic diagram of implementation of input data block segmentation in the multipath search method of the present invention; FIG. 6 is a multipath search method according to the present invention; Schematic diagram of implementation of segmentation convolution and merging; FIG. 7 is a schematic diagram of the structure of an apparatus for implementing multipath search according to the present invention. detailed description

FIG. 3 is a flowchart of a method for implementing multipath search according to the present invention. As shown in FIG. 3, the method includes the following steps: Step 300: Cut a pseudo code sequence into a preset number of subsequence segments.

Let L be the time domain length of the searched channel, Lc is the length of the CDMA local pseudocode and its length is a power of 2; and L^Lc, that is, the length of the pseudocode sequence is longer than the length of the channel. 4 is a schematic diagram of implementation of pseudo-code data block segmentation in the multipath search method of the present invention, as shown in FIG.

In this step, the pseudo code sequence c of length Lc is cut into a preset number of K subsequence segments, which are respectively c(l), c(2), ..., c(K), and each segment is N, And satisfy K*N=Lc. And N = L + Lo, where L ₀ is the smallest integer such that N = L + L _Q is an exponent of 2, that is, 2N is also an exponent of 2.

Step 301: After the end of the input signal sequence is expanded by zero, the user can be divided into a preset number of phases. Subsequence segments that overlap each other.

FIG. 5 is a schematic diagram of the implementation of input data block segmentation in the multipath search method of the present invention. As shown in FIG. 5, in this step, the length (L+Lc) input signal sequence X ends with L. Zero extension to length satisfies K*N+L+L ₀ = (K+1)*N. Then, according to FIG. 5, it is divided into a preset number of K sub-sequences which are partially overlapped with each other, which are respectively x(l), x(2), ..., x(K), and each segment has a length of 2Ν.

Step 302: Perform frequency domain search on each segment separately, and superimpose and merge the frequency domain search results of each segment.

FIG. 6 is a schematic diagram of the implementation of segmentation convolution and merging in the multipath search method of the present invention. As shown in FIG. 6, there are a total of one segment. Taking the kth segment as an example, in the kth segment, the kth input signal subsequence x(k) and the kth pseudocode subsequence c(k) are two inputs, using the existing frequency domain method. The frequency domain search is completed on the kth segment, and the length of the segment is obtained as L output sub-direction y. (k);

Finally, the result y _Q (l), y ₀ (2), y of the preset number of K segments is output separately. (K) all superimposed according to formula (2) to obtain the final frequency domain search result y ₀ :

y ₀ =y ₀ (X)+y ₀ (2)+...+y ₀ {K) (2) Compared with the existing frequency domain method:

Here, only the FFT operation on the input signal is considered, and the FFT operation of the local pseudo code can be obtained offline. Thus, the number of complex multiplications and complex additions required for the existing non-segmented frequency domain method is as follows:

Among them, the required complex multiplication method is: The FFT of the extended input signal requires complex multiplication method: tog ₂ 1⁄2; where L _fft is greater than

Lx, and the smallest integer that satisfies the exponent of 2.

The result of the input signal FFT multiplied by the result of the pseudo-code FFT requires that the complex multiplication method is: L _fft for IFFT requires the complex multiplication method to be: tog ₂ L _fft ; This adds the three parts, and the total required complex multiplication number is : L (l + l.g ₂ L ).

Similarly, the total number of complex additions required is: 2L _fft \og ₂ L _fft . If Lc> L, then only L _fft 〉 2L _c , total complex multiplication number > 2L _c (l + log ₂ 2L _c ), total recombination number > 4L _c log ₂ 2L _c .

In the frequency domain method under the segmentation method of the present invention, the required complex multiplication method and complex addition method are as follows: If Lc 〉 L , then only L _C = KN , the total complex multiplication method = 2^(1 + log ₂ 2N) = 2L _C (1 + log ₂ 2N) , total complex force method = 4 log ₂ 2N = 4L _C log ₂ 2N.

Obviously, compared with the existing non-segmented frequency domain method, the frequency domain method under the segmentation method of the present invention requires less complex operation than the non-segmented frequency domain method. Therefore, the present invention is used to split the entire convolution segment into a plurality of shorter convolution segments, and perform frequency domain segmentation search on each sub segment, further reducing the overall computational complexity; When the FFT hardware accelerator is completed, the segmentation process is used, and a small-sized FFT can be used to replace the search that originally required a large-size FFT, thereby reducing the cost.

For the method of the present invention, an apparatus for implementing multipath search is also provided, and a schematic structural diagram thereof is shown in FIG. 7, which at least includes: a pseudo code sequence segmentation module, an input signal sequence segmentation module, and a preset number of K frequency domains. The search module is a first frequency domain search module, a second frequency domain search module, a preset number of K frequency domain search modules, and a merge processing module, where

Each frequency domain search module is configured to perform frequency domain search on the kth segment according to the k pseudocode subsequence segment from the pseudo code sequence segmentation module and the kth input signal subsequence segment from the input signal sequence segmentation module, And outputting the k-th segment search result to the merge processing module; wherein k is 1 or 2, . . . or K;

a first frequency domain search module, configured to perform frequency domain on the first segment according to the first pseudo code subsequence segment from the pseudo code sequence segmentation module and the first input signal subsequence segment from the input signal sequence segmentation module Searching, and outputting the first segment search result to the merge processing module; a second frequency domain search module, configured to perform frequency domain on the second segment according to the second pseudo code subsequence segment from the pseudo code sequence segmentation module and the second input signal subsequence segment from the input signal sequence segmentation module Searching, and outputting the second segment search result to the merge processing module; a preset number of K frequency domain search modules, configured to be based on the Kth pseudocode subsequence segment from the pseudo code sequence segmentation module, and from the input signal sequence The Kth input signal subsequence segment of the segment module performs frequency domain search on the Kth segment, and outputs the Kth segment search result to the merge processing module; the merge processing module is used for frequency domain search results of each segment Perform superposition and merge to get the final frequency domain search results.

The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included. Within the scope of protection of the present invention.

Claims

Claim

A segmented frequency domain method for implementing multipath search, comprising:

Cutting the pseudo code sequence into a preset number of subsequence segments;

2. The segmentation frequency domain method according to claim 1, wherein the cutting the pseudo code sequence into a preset number of subsequence segments comprises:

The pseudo code sequence c of length Lc is cut into a preset number of K subsequence segments, respectively c(l), c(2), ..., c(K), each segment length is N, and satisfies K*N=Lc; N=L+Lo, where L ₀ is such that N=L+L. Is the smallest integer of the exponent of 2;

Where L is the time domain length of the searched channel, Lc is the length of the code division multiple access CDMA local pseudocode and its length is a power of 2, and L < Lc.

The segmentation frequency domain method according to claim 2, wherein, after the input signal sequence is zero-padded, the segmentation into a preset number of mutually overlapping sub-sequence segments includes: the input signal The length of the sequence X is (L+Lc), and the end of the input signal sequence X is zero-padded to a length that satisfies K*N+L+L. = (K+1)*N;

The input signal sequence after the zero-pad expansion is segmented into a preset number of K sub-sequences that are partially overlapped, respectively, x(l), x(2), ..., x(K), and each subsequence The length of the segment is 2Ν.

The segmentation frequency domain method according to any one of claims 1 to 3, wherein the performing frequency domain search for each segment is: using an existing non-segmented frequency domain search method .

The segmentation frequency domain method according to claim 4, wherein the superimposing and merging the frequency domain search results of each segment comprises: frequency domain search results respectively outputting a preset number of segments y. (l) , y ₀ (2), y. (K) superimposed and merged according to the following formula to get the final frequency domain search knot Fruit yo: y _D = y ₀ (i) + y ₀ (2) + - + y ₀ .

A segmented frequency domain device for implementing multipath search, comprising: a pseudo code sequence segmentation module, an input signal sequence segmentation module, a preset number of κ frequency domain search modules, and a merge processing module , among them,

The merging processing module is configured to superimpose and merge the frequency domain search results of each segment to obtain a final frequency domain search result.