CN101076007A - Method for cancelling interference realized in frequency region and used in WCDMA straight-station system - Google Patents

Abstract

The method is based on the Block least mean square (Block LMS) algorithm in time-domain and the linear correlation and linear convolution existed in the Block LMS; it uses the Fast Fourier Transform (FFT) in 1/2 overlap-save method and the calculation approach by direct multiplication at frequency domain to realize the fast correlation and fast convolution, and using an self-adaptive filter to realize the LMS algorithm at frequency-domain. It comprises four cyclic execution processes.

Description

The interference cancellation method that is used for the WCDMA direct discharging station in the frequency domain realization

Technical field

The present invention relates to a kind of interference cancellation method of the WCDMA of being used for direct discharging station in the frequency domain realization, exactly, relate to a kind of WCDMA of being used for direct discharging station the adaptive cancellation input signal interference and obtain the interference cancellation method of realizing at frequency domain of useful signal, belong to the auto-adaptive filtering technique field of radio communication.

Background technology

In communication system, run into the problem that how to detect and obtain useful signal under the high reject signal background through regular meeting, so Interference Cancellation AIC (adaptive interference cancellation) is the important component part in the communication system.After U.S. Bell laboratory in 1967 had at first proposed self-adaptive echo counteracting, the adaptive interference cancelling technology had obtained development faster.At present, there has been multiple adaptive algorithm to be applied to the Interference Cancellation filter, as based on least mean-square error LMS (least mean square) algorithm (referring to " ModifiedLMS Algorithms for Speech Processing with an Adaptive Noise Canceller ", publish in IEEE Transactions on Speech and Audio Processing, vol.6, Jul.1998, pp.338-351) and least square LS (least squares) algorithm (referring to " Multichannel Recursive-Least-SquaresAlgorithms and Fast-Transversal-Filter Algorithms for Active Noise Control andSound Reproduction Systems ", publish in IEEE Transactions On Speech And AudioProcessing, vol.8,2000, pp.606-618.) etc., these algorithms have been applied to a plurality of fields.

In numerous improvement LMS algorithms, to the adaptive block least mean-square error Block LMS algorithm that time-domain signal is operated, its self adaptation of filter process is based on that the signal data piece carries out, and these are different with traditional LMS algorithm based on symbol.Its concrete grammar is: the time-domain signal data flow u (n) of input is exported from L bar branch road through behind the serial to parallel conversion, wherein the signal data of every branch road is all formed a data block with the tap number M of filter as block length, and then k time-domain signal data block is A ^TAnd A (k), ^T(k)=[u (kL), u (kL+1) ..., u (kL+L-1)], in the formula, u (kL+i) is the vector representation of k input signal data piece at i+1 branch road, and the span of branch road sequence number i is: [0, L-1], then have u (kL+i)=[u (kL+i), u (kL+i-1) ..., u (kL+i-M+1)] ^T

After more above-mentioned input signal data piece being passed through filter, the output signal that obtains is: $y(\mathrm{kL}+i)={\hat{w}}^T\left(k\right)u(\mathrm{kL}+i)=\underset{j=0}{\overset{M-1}{\mathrm{\Σ}}}{\hat{w}}_j\left(k\right)u(\mathrm{kL}+i-j),$ In the formula, Be k the pairing filter tap coefficients of signal data piece

Transposition, and $\hat{w}\left(k\right)={[{\hat{w}}_0\left(k\right),{\hat{w}}_1\left(k\right),\·\·\·,{\hat{w}}_{M-1}\left(k\right)]}^T.$ Because people normally with error signal e (kL+i) (in technical scheme of the present invention, this error signal is referred to as useful signal) be defined as desired signal r (kL+i) (in technical scheme of the present invention, this desired signal is referred to as disturbed signal) and filter output signal y (kL+i) (in technical scheme of the present invention, this output signal is referred to as the estimated value of interference) poor, i.e. error signal e (kL+i)=r (kL+i)-y (kL+i).According to the LMS algorithm, for making the mean square error minimum of error signal, then the tap coefficient in the piece LMS algorithm is updated to: $\hat{w}(k+1)=\hat{w}\left(k\right)+\mathrm{\μ}\underset{i=0}{\overset{L-1}{\mathrm{\Σ}}}u(\mathrm{kL}+i)e(\mathrm{kL}+i),$ Wherein μ is an iteration step length.

In interference cancellation systems, people can utilize the above-mentioned various adaptive algorithms of mentioning to carry out Interference Cancellation.But when disturbing time-delay very big, the tap number of sef-adapting filter must be quite a lot of, can be enough to offset this interference.Especially in the WCDMA system, because each signal data bit spreads to 128 chips, and, each chip is sampled as a plurality of sample values again in digital communication system, therefore, the memory span of disturbing in several microseconds just may be crossed over hundreds of sample values, and so long memory span must use the very many filters of number of taps can offset interference.In this case, if carry out the LMS algorithm, certainly will cause extremely complicated difficulty in computation in time domain.Though because time averaging effect, piece LMS algorithm can obtain more accurate gradient vector and estimate with respect to traditional LMS algorithm based on symbol; But if block LMS algorithm realizes in time domain that still under the considerable situation of filter tap number, it will be that inevitably this just certainly will influence the real-time processing and the realization of signal that magnanimity is calculated.Therefore, how the method is improved and just become scientific and technical personnel's a research focus in the industry.

Summary of the invention

In view of this, the interference cancellation method of realizing at frequency domain that the purpose of this invention is to provide a kind of WCDMA direct discharging station just provides a kind of and adopts sef-adapting filter and the interference of offset input signal and obtain the implementation method that the frequency domain interference of useful signal is offset.This method not only extracts useful signal effectively, and greatly reduces the workload and the complexity of calculating.

In order to achieve the above object, the invention provides a kind of WCDMA of being used for direct discharging station the adaptive cancellation input signal interference and obtain the implementation method that the frequency domain interference of useful signal is offset, it is characterized in that: this method is based on data block least mean-square error (the Block LMS in the time domain, block least meansquare) there is the process of linear correlation and linear convolution in computational methods and this piece LMS algorithm, fast fourier transform FFT (fast fourier transforms) by 1/2 overlap-save method realizes fast correlation and fast convolution at frequency domain with the account form that directly multiplies each other, and utilizes sef-adapting filter to realize the LMS algorithm at frequency domain; Comprise the operating procedure that following circulation is carried out:

(1) the frequency domain tap coefficient of sef-adapting filter is done the initialization setting, and the time domain input signal of this filter is N point discrete fast Fourier transform FFT handles, make it be converted to frequency-region signal, as the input signal of sef-adapting filter; Wherein N is 2 times of tap number M of this filter;

(2) frequency-region signal with input carries out the adaptive-filtering processing by sef-adapting filter, and the output signal of this filter is carried out invert fast fourier transformation IFFT (inverse fast fourier transforms) handle, make it be converted to time-domain signal, as the estimated value of disturbing;

(3) difference between the time-domain signal of disturbed signal of calculating and filter output is as useful signal; Produce the frequency domain value of useful signal again;

(4) utilizing frequency-region signal to carry out least mean-square error LMS calculates, promptly the frequency domain value according to useful signal and filter input signal upgrades filter tap coefficients, so that returning when carrying out above-mentioned steps (2), use tap coefficient after this renewals to continue to carry out relevant adaptive-filtering processing again and again from the new frequency domain input signal of step (1).

The present invention is a kind of interference cancellation method in the frequency domain realization that obtains useful signal of the WCDMA of being used for direct discharging station, this method is utilized the thinking of frequency domain fast fourier transform, the adaptive-filtering process of time domain is transformed in the frequency domain realizes, thus the interference in the adaptive cancellation input signal.The present invention not only extracts useful signal effectively, has guaranteed convergence, and compares with the interference cancellation method of time domain, greatly reduces algorithm complex.In a word, the efficient height of the inventive method, speed are fast, and, can obviously improve the power spectral density and the planisphere performance of system, under the condition that reduces algorithm complex greatly, can guarantee the validity of algorithm, have good engineering application and application prospect.

Description of drawings

Fig. 1 is the schematic diagram of the signals transmission of WCDMA direct discharging station.

Fig. 2 is the interference of the present invention's adaptive cancellation input signal of being used for the WCDMA direct discharging station and obtain the frequency domain interference counteracting method flow diagram of useful signal.

Fig. 3 is the inventive method four carrier signals of being used for the WCDMA direct discharging station at signal interference ratio under-10dB the condition, and the performance that Interference Cancellation improves power amplifier power output spectrum density (PSD) is schematic diagram relatively.

Fig. 4 is that four carrier signals that the inventive method is used for the WCDMA direct discharging station are under the 0dB condition at signal interference ratio, and the performance that Interference Cancellation improves power amplifier output power spectrum density performance (PSD) is schematic diagram relatively.

Fig. 5 is that four carrier signals that the inventive method is used for the WCDMA direct discharging station are under the 5dB condition at signal interference ratio, and the performance that Interference Cancellation improves power amplifier output power spectrum density performance (PSD) is schematic diagram relatively.

Fig. 6 (A), (B) be respectively do not adopt the inventive method and adopt the inventive method for four carrier signals of WCDMA direct discharging station at signal interference ratio under-10dB the condition, two planisphere performances of power amplifier output are schematic diagram relatively.

Fig. 7 (A), (B) are respectively that not adopt the inventive method and adopt the inventive method be under the 0dB condition for four carrier signals of WCDMA direct discharging station at signal interference ratio, and two planisphere performances of power amplifier output are schematic diagrames relatively.

Fig. 8 (A), (B) are respectively that not adopt the inventive method and adopt the inventive method be under the 5dB condition for four carrier signals of WCDMA direct discharging station at signal interference ratio, and two planisphere performances of power amplifier output are schematic diagrames relatively.

Embodiment

For making the purpose, technical solutions and advantages of the present invention clearer, the present invention is described in further detail below in conjunction with accompanying drawing.

The inventive method is based on the process that has linear correlation and linear convolution in data block least mean-square error Block LMS computational methods in the time domain and this piece LMS algorithm, FFT by 1/2 overlap-save method, realize fast correlation and fast convolution at frequency domain with the account form that directly multiplies each other, utilize sef-adapting filter to realize the LMS algorithm at frequency domain.

Referring to Fig. 1 and Fig. 2, specifically introduce the present invention below respectively and be used for the method for WCDMA direct discharging station Interference Cancellation sef-adapting filter and four operating procedures that circulation is carried out ((1) that Fig. 2 goes out with the fine line structure, (2), (3), (4) four corresponding respective phases of operation of square frame difference) thereof:

(1) the frequency domain tap coefficient of sef-adapting filter is done the initialization setting, and the time domain input signal of this filter is N point discrete fast Fourier transform FFT handles, make it be converted to frequency-region signal, as the input signal of sef-adapting filter; Wherein N is 2 times of tap number M of this filter.

The concrete operations content of this step is:

(11) the frequency domain tap coefficient of initialization sef-adapting filter: the tap number that the Interference Cancellation sef-adapting filter is set is M, and the feedback signal that this M numerical value is greater than interference signal and power amplifier output arrives filter delay time constantly between the two; Because adopt the FFT (time-domain signal being become frequency-region signal) of 1/2 overlap-save method by FFT, M 0 should filled to M time domain tap coefficient of this filter when the initialization process thereafter, to constitute the time-domain signal initial value of N=2M, then this N point is made FFT and calculate, the initial tap coefficient that obtains frequency-region signal is $\hat{w}\left(k\right)=\mathrm{FFT}\left[\begin{array}{c}\hat{w}\left(k\right)\\ 0\end{array}\right],$ In the formula,

Be the initial tap coefficient of time-domain signal,

It is the initial tap coefficient of frequency-region signal; And this initialization step (11) only carried out before this method begins to import first data block, and only carried out once; For the second time and subsequent each tap coefficient then all obtain by subsequently adaptive-filtering renewal process; When promptly returning step (2) by step (4), the frequency domain input data that sef-adapting filter is new exist, and are directly provided by operating procedure (12).

(12), and, form continuous data block with the unit of M signal as data block with the time domain input signal of power amplifier feedback signal u (n) as filter, again with two continuous data block cascades, then two data blocks of this cascade are N point FFT, obtain frequency-region signal U (k), then have:

Wherein, M is the tap number of sef-adapting filter, N=2M, diag{a ₁, a ₂..., a _nRepresent with a ₁, a ₂..., a _nDiagonal matrix as element on the leading diagonal.

In order to realize the effect of Interference Cancellation, the input signal u (n) of filter should be relevant with the interference signal in the disturbed signal, and uncorrelated with useful signal.Owing to disturb to the signal (referring to Fig. 1) of power amplifier output, therefore, choose the input signal u (n) of the output of power amplifier in the inventive method as filter through dissemination channel.

(2) frequency-region signal with input carries out the adaptive-filtering processing by sef-adapting filter, and the output signal of this filter is carried out invert fast fourier transformation IFFT (inverse fast fourier transforms) handle, make it be converted to time-domain signal, as the estimated value of disturbing.

The concrete operations content of this step is:

(21) on traditional time domain, the time-domain signal of input should with the time domain tap coefficient of filter convolution mutually.Because the present invention has been transformed into frequency domain with input signal and filter tap coefficients, therefore on frequency domain, just can use the input signal U (k) of frequency domain and the filter tap coefficients of frequency domain Directly multiply each other, obtain the filter output signal of frequency domain $Y\left(k\right)=U\left(k\right)\hat{W}\left(k\right),$ Thereby realize replacing the convolution of time-domain signal, simplify amount of calculation greatly with the product of frequency domain.

(22) frequency domain output signal Y (k) is contrary fast fourier transform IFFT and handles, the frequency-region signal that is about to filter output is transformed to time-domain signal.

(23) only keep the regulation of M thereafter useful time domain data according to 1/2 overlap-save method, the IFFT result of above-mentioned steps (22) is only kept M useful time domain data thereafter, obtain k time-domain data blocks y ^T(k), promptly $y^T\left(k\right)=[y\left(\mathrm{kM}\right),\·\·\·,y(\mathrm{kM}+M-1)]=\mathrm{IFFT}\left[U\left(k\right)\hat{W}\left(k\right)\right]$ Back M data; Then yT (k) is carried out the time domain output signal y (k) that matrix transpose obtains filter, this data block y (k) promptly is the estimated value to interference signal among k the data block r (k) of disturbed signal.

(3) difference between the time-domain signal of disturbed signal of calculating and filter output is as useful signal; Produce the frequency domain value of useful signal again.

The concrete operations content of this step is:

(31) be that block unit is formed each data block with disturbed signal r (n) with M signal, k the data block r (k) in the then disturbed signal is: r (k)=[r (kM), r (kM+1) ..., r (kM+M-1)] ^T

(32) calculate both poor of the interference signal estimated value y (k) of k data block r (k) and the middle institute of above-mentioned steps (23) corresponding data piece in the disturbed signal, be k useful signal data block d (k)=[d (kM) behind the Interference Cancellation,, d (kM+M-1)] ^T=r (k)-y (k); And with this k useful signal data block d (k) as the useful signal of removing interference, export to power amplifier.

(33), for keeping consistency,, carry out FFT then, the useful signal data block of calculating frequency domain at the preceding interpolation M of useful signal data block d (k) 0 because the present invention has abandoned a preceding M data value when obtaining y (k) in above-mentioned steps (23) $D\left(k\right)=\mathrm{FFT}\left[\begin{array}{c}0\\ d\left(k\right)\end{array}\right].$

(4) utilizing frequency-region signal to carry out least mean-square error LMS calculates, promptly the frequency domain value according to useful signal and filter input signal upgrades filter tap coefficients, so that returning when carrying out above-mentioned steps (2), use tap coefficient after this renewals to continue to carry out relevant adaptive-filtering processing again and again from the new frequency domain input signal of step (1).

The concrete operations content of this step is:

(41) utilize frequency domain LMS algorithm, the Matrix Conjugate transposition U of k the data block frequency domain input signal U (k) that step (12) is obtained ^H(k) the frequency domain useful data information D (k) that obtains with step (33) multiplies each other, and obtains both product T (k)=U ^H(k) D (k); Then, according to 1/2 overlap-save method, handle acquisition time-domain signal Φ (k): Φ (k)=IFFT[U by IFFT ^H(k) D (k)] preceding M data;

(42) for initial value $\hat{W}\left(k\right)=\mathrm{FFT}\left[\begin{array}{c}\hat{w}\left(k\right)\\ 0\end{array}\right]$ The form of back benefit M individual 0 is corresponding, after M individual 0 is filled in Φ (k) back, carries out FFT and handles, promptly

(43) upgrade filter tap coefficients, the filter tap coefficients after obtaining upgrading at frequency domain

Multiply each other for use in k+1 the block of frequency domain data of feeding back with power amplifier output in the step (2), and $\hat{w}(k+1)=\hat{w}\left(k\right)+\mathrm{\μFFT}\left[\begin{array}{c}\mathrm{\Φ}\left(k\right)\\ 0\end{array}\right],$ In the formula,

Be this filtering and k block of frequency domain data employed tap coefficient that multiplies each other,

Be filtering next time and k+1 block of frequency domain data employed tap coefficient that multiplies each other; μ is an iteration

Iteration step length in the process is used to determine the mean-square value E[|d (k) of useful signal | ²] converge to the speed of best mean-square value and the accuracy of convergency value.The numerical value of this iteration step length μ is tested to compromise and is chosen by adjusting its numerical value in actual mechanical process, and its rule is that the numerical value of μ is big more, and tap coefficient convergence is to fast more near the speed of optimal value, but convergency value is inaccurate more, and vice versa.

Referring to Fig. 3～Fig. 8, introduce an experimental example of the inventive method, and will adopt the inventive method to realize the system of Interference Cancellation and do not have the systematic function of Interference Cancellation to compare at frequency domain.Suppose that signal transmission delay is 6 microseconds between two antennas in the four carrier wave WCDMA systems of a 20MHz bandwidth, and suppose that the channel between these two antennas is two footpath fading channels, iteration step length μ=0.0001.Four carrier signal sample rates are 16, and the sef-adapting filter tap length is 512, and FFT length is 1024.Power amplifier PA (power amplifier) receives model for dimension, and signal interference ratio is defined as the ratio of reception antenna end chip signal power and interference power.Compare as performance index with the stopband decline dB value of power spectral density PSD (power spectral density) and the error vector magnitude EVM (error vector magnitude) of planisphere.Thick dashed line among Fig. 3～Fig. 5 is represented the power spectral density of information source, fine dotted line represents to have interference signal directly to pass through the power spectral density of PA, power spectral density after solid line represents have interference signal through AIC of the present invention and PA processing, chain-dotted line represents not have the power spectral density of interference signals through PA.

Below two forms list file names with when direct discharging station is input as other carrier numbers, use the performance index comparable situation of AIC of the present invention (adaptive interference cancellation) method.

Table 1 is that the present invention is used for the WCDMA direct discharging station, and the power spectral density performance that each carrier signal uses frequency domain interference to offset under different signal to noise ratio conditions is relatively tabulated:

Carrier number	Signal interference ratio (dB)	PSD figure index (the dB value that stopband descends)
					Directly through PA	Through AIC of the present invention, PA	Information source
		Single carrier	Noiseless	27					42
-10	22				27
			0	26			27
5	26				27
			Two carrier waves	Noiseless		28			40
-10	12	20
				0	19	20
5	19	20
				Three carrier waves	Noiseless	25		38
-10	10	19
			0		18	19
5	18	19

Four carrier waves	Noiseless	23		38
					-10	10	17
	0	16	17
					5	16	17

Referring to Fig. 3～Fig. 5 and table 1, can obtain such conclusion: adopt frequency domain interference counteracting method of the present invention under low signal interference ratio situation, obviously to improve the PSD performance.

Table 2 is that the present invention is used for the WCDMA direct discharging station, and the planisphere performance that each carrier signal uses frequency domain interference to offset under different signal to noise ratio conditions is relatively tabulated:

Carrier number	Signal interference ratio (dB)	Planisphere index (statistics EVM%)
				Directly through PA	Through AIC of the present invention, PA
		Single carrier	Noiseless			7.6164
-10	57.5177			20.1308
			0		76.6166	20.4842
5	35.5381			20.1992
			Two carrier waves		Noiseless	4.6440
-10	141.9632	16.6292
				0	139.6261	16.3564
5	37.5836	16.7826
				Three carrier waves	Noiseless	6.3061
-10	138.7951	17.5369
			0		134.1968	17.8611
5	37.3231	17.5417
			Four carrier waves		Noiseless	7.8442
-10	44.4481	18.5394
				0	70.7926	18.5580
5	34.6272	18.9823

Referring to Fig. 6～Fig. 8 and table 2, can obtain such conclusion: adopt frequency domain interference counteracting method of the present invention under any signal interference ratio situation, obviously to improve the planisphere performance.

Algorithm complex to the frequency domain interference counteracting method that adopts among the present invention and traditional time domain interference cancellation method compares below.When adopting hardware to realize, computation complexity often is decided by the number of times of multiplying, multiplication number that therefore can more above-mentioned two kinds of methods.For the time domain interference cancellation method that M filter tap arranged,, then need 2M altogether because each data block has M data ²Inferior multiplying; And for the frequency domain interference counteracting method that M filter tap arranged, total multiplication number of times is 10M log ₂M+26M.The algorithm complex ratio of frequency domain interference counteracting method and time domain interference cancellation method is about (5log so ₂M+13)/M.Therefore, when filter tap coefficients was very big, the computation complexity of frequency domain interference counteracting method will be well below the time domain interference cancellation method.In a word, frequency domain interference counteracting method of the present invention can obviously improve the power spectral density and the planisphere performance of system, has therefore guaranteed the validity of algorithm under the condition that reduces algorithm complex greatly, has good engineering application.

Claims (8)

1, a kind of interference of adaptive cancellation input signal of the WCDMA of being used for direct discharging station and obtain the interference cancellation method of realizing at frequency domain of useful signal, it is characterized in that: this method is based on the process that has linear correlation and linear convolution in the data block least mean-square error Block LMS computational methods in the time domain and this piece LMS algorithm, fast fourier transform FFT by 1/2 overlap-save method, realize fast correlation and fast convolution at frequency domain with the account form that directly multiplies each other, utilize sef-adapting filter to realize the LMS algorithm at frequency domain; Comprise the operating procedure that following circulation is carried out:

(1) the frequency domain tap coefficient of sef-adapting filter is done the initialization setting, and the time domain input signal of this filter is N point discrete fast Fourier transform FFT handles, make it be converted to frequency-region signal, as the input signal of sef-adapting filter; Wherein N is 2 times of tap number M of this filter;

(2) frequency-region signal with input carries out the adaptive-filtering processing by sef-adapting filter, and the output signal of this filter is carried out invert fast fourier transformation IFFT handle, and makes it be converted to time-domain signal, as the estimated value of disturbing;

(3) difference between the time-domain signal of disturbed signal of calculating and filter output is as useful signal; Produce the frequency domain value of useful signal again;

(4) utilizing frequency-region signal to carry out least mean-square error LMS calculates, promptly the frequency domain value according to useful signal and filter input signal upgrades filter tap coefficients, so that returning when carrying out above-mentioned steps (2), use tap coefficient after this renewals to continue to carry out relevant adaptive-filtering processing again and again from the new frequency domain input signal of step (1).

2, the interference cancellation method of realizing at frequency domain according to claim 1 is characterized in that: the input signal u (n) of described filter is relevant with interference signal in the disturbed signal, and uncorrelated with useful signal; And described interference is the signal behind power amplifier output and the process dissemination channel, therefore chooses the input signal u (n) of the output of power amplifier as sef-adapting filter, so that realize the effect that this interference is cancelled.

3, the interference cancellation method of realizing at frequency domain according to claim 1 is characterized in that described step (1) further comprises following content of operation:

(11) the frequency domain tap coefficient of initialization sef-adapting filter: the tap number that the Interference Cancellation sef-adapting filter is set is M, and the feedback signal that this M numerical value is greater than interference signal and power amplifier output arrives filter delay time constantly between the two; Because adopt the FFT of 1/2 overlap-save method, M 0 should filled to M time domain tap coefficient of this filter when the initialization process thereafter, to constitute the time-domain signal initial value of N=2M, then this N point to be made FFT and calculate, the initial tap coefficient that obtains frequency-region signal is $\hat{W}\left(k\right)=\mathrm{FFT}\left[\begin{array}{c}\hat{w}\left(k\right)\\ 0\end{array}\right],$ In the formula,

Be the initial tap coefficient of time-domain signal,

It is the initial tap coefficient of frequency-region signal;

(12), and, form continuous data block with the unit of M signal as data block with the time domain input signal of power amplifier feedback signal u (n) as filter, again with two continuous data block cascades, then two data blocks of this cascade are N point FFT, obtain frequency-region signal U (k), then have:

Wherein, M is the tap number of sef-adapting filter, N=2M, diag{a ₁, a ₂..., a _nRepresent with a ₁, a ₂..., a _nDiagonal matrix as element on the leading diagonal.

4, according to claim 1 or the 3 described interference cancellation methods of realizing at frequency domain, it is characterized in that: the operation of described step (11) must begin input signal in this method to be finished in the past in advance, and only carried out once; For the second time and subsequent each tap coefficient then all obtain by subsequently adaptive-filtering renewal process; When promptly returning step (2) by step (4), the frequency domain input data that sef-adapting filter is new exist, and are directly provided by operating procedure (12).

5, the interference cancellation method of realizing at frequency domain according to claim 1 is characterized in that described step (2) further comprises following content of operation:

(21) with the input signal U (k) of frequency domain and the filter tap coefficients of frequency domain Directly multiply each other, obtain the frequency domain output signal of sef-adapting filter $Y\left(k\right)=U\left(k\right)\hat{W}\left(k\right),$ To substitute the process of convolution of time-domain signal;

(22) frequency domain output signal Y (k) being contrary fast fourier transform IFFT handles;

(23) according to the regulation of 1/2 overlap-save method, the IFFT result of above-mentioned steps (22) is only kept M useful time domain data thereafter, obtain y ^T(k), promptly

$y^T\left(k\right)=[y\left(\mathrm{kM}\right),\·\·\·,y(\mathrm{kM}+M-1)]=\mathrm{IFFT}\left[U\left(k\right)\hat{W}\left(k\right)\right]$ Back M data; Then to y ^T(k) carry out the time domain output signal y (k) that matrix transpose obtains filter, this data block y (k) promptly is the estimated value to interference signal among k the data block r (k) of disturbed signal.

6, the interference cancellation method of realizing at frequency domain according to claim 1 or 5 is characterized in that described step (3) further comprises following content of operation:

(31) be block unit with M signal, disturbed signal r (n) is formed each data block, k the data block r (k) in the then disturbed signal is: r (k)=[r (kM), r (kM+1) ..., r (kM+M-1)] ^T

(32) calculate both poor of the interference signal estimated value y (k) of k data block r (k) and the middle institute of above-mentioned steps (23) corresponding data piece in the disturbed signal, obtain k useful signal data block d (k)=[d (kM) behind the Interference Cancellation,, d (kM+M-1)] ^T=r (k)-y (k); And with this k useful signal data block d (k) as the useful signal of removing interference, export to power amplifier;

(33) because above-mentioned steps (23) has abandoned a preceding M data value when obtaining y (k), individual 0 at the preceding interpolation M of useful signal data block d (k) for keeping consistency, carry out FFT then, calculate the useful signal data block of frequency domain $D\left(k\right)=\mathrm{FFT}\left[\begin{array}{c}0\\ d\left(k\right)\end{array}\right].$

7, according to claim 1 or the 3 or 6 described interference cancellation methods of realizing at frequency domain, it is characterized in that described step (4) further comprises following content of operation:

(41) utilize frequency domain LMS algorithm, the Matrix Conjugate transposition U of k the data block frequency domain input signal U (k) that step (12) is obtained ^H(k) the frequency domain useful data information D (k) that obtains with step (33) multiplies each other, and obtains both product T (k)=U ^H(k) D (k); Then, according to 1/2 overlap-save method, handle acquisition time-domain signal Φ (k): Φ (k)=IFFT[U by IFFT ^H(k) D (k)] preceding M data;

(42) for initial value $\hat{W}\left(k\right)=\mathrm{FFT}\left[\begin{array}{c}\hat{w}\left(k\right)\\ 0\end{array}\right]$ The form of back benefit M individual 0 is corresponding, after M individual 0 is filled in Φ (k) back, carries out FFT and handles, promptly $\mathrm{FFT}\begin{array}{c}\left[\begin{array}{c}\mathrm{\Φ}\left(k\right)\\ 0\end{array}\right]\end{array};$

(43) upgrade filter tap coefficients, the filter tap coefficients after obtaining upgrading at frequency domain

Multiply each other for use in k+1 the block of frequency domain data of feeding back with power amplifier output in the step (2), and $\hat{W}(k+1)=\hat{W}\left(k\right)+\mathrm{\μFFT}\left[\begin{array}{c}\mathrm{\Φ}\left(k\right)\\ 0\end{array}\right],$ In the formula,

Be this filtering and k block of frequency domain data employed tap coefficient that multiplies each other, Be filtering next time and k+1 block of frequency domain data employed tap coefficient that multiplies each other; μ is an iteration

Iteration step length in the process is used to determine the mean-square value E[|d (k) of useful signal | ²] converge to the speed of best mean-square value and the accuracy of convergency value.

8, the interference cancellation method of realizing at frequency domain according to claim 7, it is characterized in that: the numerical value of described iteration step length μ is tested to compromise and is chosen by adjusting its numerical value in actual mechanical process, its rule is that the numerical value of μ is big more, tap coefficient convergence is to fast more near the speed of optimal value, but convergency value is inaccurate more, and vice versa.

Images