Signal processing device and signal processing method

Abstract

One aspect of the present invention comprises: a Fourier transform processing unit that transforms an input signal into a frequency domain signal for each block of size N; a multiplication processing unit that performs filter processing on the frequency domain signal transformed by the Fourier transform processing unit using a filter weighting coefficient; an inverse Fourier transform processing unit that performs inverse Fourier transform processing on the frequency domain signal on which the filter processing has been performed by the multiplication processing unit; and an output signal selection unit that, among output symbols of size N obtained by the inverse Fourier transform processing unit performing inverse Fourier transform processing, stores N / 2 + Δ symbols centered on the N / 2 sized output symbol from the beginning, and outputs a signal in which output symbols other than the stored output symbols are discarded.

Description

Technical Field

[0001] The present invention relates to the technology of a signal processing apparatus and a signal processing method.

Background Art

[0002] With the start of recent 5G (5th Generation) services, the development of high-definition video service delivery, and the development of IoT (Internet of Things) services, etc., the communication traffic flowing through the optical network has been increasing year by year. As countermeasures in the optical network against the increasing communication traffic demand, for example, without changing the structure of the optical fiber as a transmission line, countermeasures such as enhancing the functionality of the optical communication system apparatus installed at the end stations of the optical network and introducing optical amplifiers and optical switches have been taken.

[0003] The optical fiber that forms the basis of the current large-capacity optical network uses single-mode fiber (SMF) except for local networks for short distances such as LAN (Local Area Network). The single-mode fiber has a single core that serves as a passage for optical signals in the cladding, and is an optical fiber designed to allow only single-mode propagation in wavelength bands such as the C band and L band used in large-capacity long-distance optical networks. Thereby, a large-capacity long-distance optical network that can stably transfer information reaching several terabits per second over a long distance has been realized.

[0004] In the above optical network, digital coherent transmission technology that uses digital signal processing technology and coherent transceiver technology has been commercially introduced into optical transmission apparatuses of the order of 100 gigabits per second. Digital coherent transmission technology is a technology that combines a coherent reception method and ultra-high-speed digital signal processing. The coherent reception method is a reception method that detects the interference light between the optical signal and the local oscillation light on the reception side. Ultra-high-speed digital signal processing is a process that reproduces the envelope waveform of the optical signal in the digital domain and performs equalization of waveform distortion generated in the transmission line and transceivers.

[0005] By employing digital coherent transmission technology, waveform distortion can be effectively eliminated based on models of the underlying physical mechanisms, resulting in the realization of compact, inexpensive, and low-power optical transceivers. The advent of digital coherent transmission technology has not only improved reception sensitivity during optical transmission in high-capacity optical networks, but has also dramatically improved information transmission efficiency by embedding information in the amplitude, phase, and polarization of the optical carrier wave.

[0006] A more specific example of a transmission method utilizing digital coherent transmission technology, which encodes information onto polarization in optical transmission systems, is polarization multiplexing (PDC) optical transmission using two orthogonal polarization modes on a single-mode fiber. In PDC, different information can be encoded onto orthogonal polarizations. When PDC is performed, orthogonal polarizations mix in a complex manner within the optical transmission path, and the orthogonal axes of the polarization modes fluctuate rapidly. Therefore, it is difficult to track such polarizations using optical devices.

[0007] Therefore, a receiver compatible with polarization diversity structures receives mixed polarization multiplexed optical signals, converts the received polarization multiplexed optical signals into digital signals, and separates them using digital signal processing. This process can be modeled as a 2x2 MIMO (Multiple-Input Multiple-Output) system used in wireless communication systems. This makes it possible to extract information for each polarization from the separated signals, establishing communication between the transmitter and receiver.

[0008] Another example of a transmission method using digital coherent transmission technology is mode-multiplexing optical transmission, which uses multiple spatial modes (hereinafter also referred to as "modes") in a multimode optical fiber. In mode-multiplexing optical transmission, for example, a fiber with a wider core diameter compared to a single-mode fiber is used as the transmission medium. This makes it possible to excite multiple modes even in existing wavelength bands such as the C band, and transmit different information on each mode.

[0009] In mode-multiplexed optical transmission, as in polarization-multiplexed optical transmission, the mode-multiplexed optical signals are complexly mixed as they propagate through multimode optical fibers. A receiver compatible with mode diversity structures receives the mixed mode-multiplexed optical signals, converts the received mode-multiplexed optical signals into digital signals, and separates them using MIMO-type signal processing on a scale corresponding to the number of excited modes.

[0010] As a more concrete example, consider a multimode fiber that excites two LP (Linearly Polarized) modes. In a multimode fiber for 2LP modes, the ground mode, LP01 mode, and the higher-order mode, LP11 mode, are excited. Furthermore, by utilizing the two degenerate modes of LP11 mode (referred to as LP11a and LP11b, respectively) and the polarization modes of each mode (referred to as X-polarization and Y-polarization, respectively), a multimode fiber for 2LP modes can carry different information in a total of six spatial modes: LP01X, LP01Y, LP11aX, LP11aY, LP11bX, and LP11bY. Therefore, if we ignore the nonlinear optical effects of optical fibers, in principle, a multimode fiber for 2LP modes can achieve three times the transmission capacity of existing single-mode fibers.

[0011] Thus, in spatial division multiplexing transmission technology, which transmits different and independent information to the propagating light in each spatial mode of the multimode optical fiber, an improvement in transmission capacity per optical fiber can be expected by the number of spatial modes excited. Research and development by related organizations toward practical application is a global trend in order to realize future high-capacity optical backbone networks.

[0012] MIMO signal processing requires not only coupling between spatial modes but also a function to compensate for phenomena caused by delay differences between signal pulses in the time domain, known as dispersion. The dispersion referred to here is a phenomenon arising from the group delay difference between waveguide modes, and refers to, for example, polarization mode dispersion that occurs in single-mode optical fibers and mode dispersion that occurs in multimode optical fibers.

[0013] Generally, dispersion has a cumulative characteristic depending on the transmission distance. Therefore, MIMO-type signal processing for optical signals transmitted over long distances requires the application of MIMO-type signal processing with a finite impulse response (FIR) that has a number of multipliers (number of taps) that sufficiently encompasses the temporal spread of the signal pulse due to dispersion (this processing will be referred to as MIMO-FIR-type signal processing from now on). In other words, the required number of taps increases with transmission distance, leading to an increase in the size of the signal processing circuit.

[0014] As an effective method for reducing the signal processing circuit size for the above-mentioned MIMO-FIR signal processing, frequency-domain MIMO-FIR signal processing, which performs time-domain signal processing in the frequency domain, is known (see Non-Patent Documents 1 and 2). Frequency-domain MIMO-FIR signal processing is a processing method that effectively reduces the signal processing size of MIMO-FIR signal processing by applying processing via the Fast Fourier Transform, based on the fact that cyclic convolution is equivalent to element-wise product in the frequency domain. MIMO-FIR signal processing makes it possible to separate spatial mode coupling including polarization and to compensate for dispersion generated in the transmission fiber all at once. [Prior art documents] [Non-patent literature]

[0015] [Non-Patent Document 1] Mansour, D., & Gray, A. (1982). Unconstrained frequency-domain adaptive filter. IEEE Transactions on Acoustics, Speech, and Signal Processing, 30(5), 726-734. [Non-Patent Document 2] Md. Saifuddin Faruk and Kazuro Kikuchi, "Adaptive frequency-domain equalization in digital coherent optical receivers," Opt. Express 19, 12789-12798 (2011) [Overview of the project] [Problems that the invention aims to solve]

[0016] Conventional frequency-domain MIMO-FIR signal processing has employed a block-processing type overlap preservation method (for example, the method described in Non-Patent Document 2). The overlap rate referred to here is the ratio of the number of input signal samples that overlap between block number k and block number k+1 in the processing of the overlap preservation method.

[0017] In overlap preservation processing, if the length of one block in the block processing is N (where N is a natural number), then generally, N is a power of 2 that is sufficiently large to efficiently perform the Fourier transform processing using the Fast Fourier Transform. Furthermore, frequency domain MIMO-FIR signal processing based on overlap preservation is usually operated with an overlap rate of 50%, and N is set to N = 2PM (hereinafter referred to as "Equation 1"). Here, M is the number of output symbols per block, and P is the oversampling rate of the input signal. The overlap rate of 50% is set as the rate under which no inter-block interference occurs in the updated output signal and filter weight coefficients related to overlap preservation.

[0018] FIG. 9 is a diagram showing the correspondence between the input signal sequence and the output signal sequence of the conventional frequency-domain MIMO-FIR type signal processing with an overlap rate of 50%. For the k-th block, the latter half of N / 2 symbols is stored as the output signal. Similarly, for the (k + 1)-th block, the latter half of N / 2 symbols is stored as the output signal. By concatenating the output signals obtained from these consecutive blocks, a continuous output signal sequence is obtained.

[0019] The tap length L defining the FIR type filter processing is related to L = PM (hereinafter referred to as "Equation 2") and is generally set according to the memory length of the transmission path channel to be compensated. In the case of optical fiber communication under consideration here, the tap length L is set so that it can sufficiently compensate for the impulse response spread caused by the dispersion phenomenon. Another requirement for the tap length L is the condition that N is a power of 2 in consideration of (Equation 1).

[0020] In view of the above-described setting conditions, when applying the frequency-domain MIMO-FIR type signal processing based on the fast Fourier transform processing to a long-distance optical fiber communication system or the like, a discrete increase in the tap length L occurs. As an example, consider the case where the tap length required for compensating the dispersion phenomenon at a distance D1 = D is L, and the tap length required for compensating the dispersion phenomenon at a distance D2 = D + ΔD is L + ΔL.

[0021] Based on (Equation 1), the required block length N(D2) at the distance D2 is set to N(D2) = 2(L + ΔL) (hereinafter referred to as "Equation 3") when ΔL < L holds. However, due to the requirement that the block length of the fast Fourier transform processing is a power of 2, in actual operation, it is set to N(D2) = 4L (hereinafter referred to as "Equation 4").

[0022] Therefore, in the conventional frequency-domain MIMO-FIR type signal processing with an overlap rate of 50%, discrete (discontinuous) increases occur in the block length N and the tap length L, which are processing parameters, from distance D1 to distance D2. That is, the required conditions for the processing parameters are not consistent between the physical dispersion phenomenon and the computational efficiency for fast Fourier transform processing. In particular, the output symbol M obtained by the conventional frequency-domain MIMO-FIR type signal processing with an overlap rate of 50% is set too small, resulting in a problem of reduced computational efficiency.

[0023] In view of the above circumstances, an object of the present invention is to provide a technique capable of reducing the amount of calculation.

Means for Solving the Problems

[0024] One aspect of the present invention is a signal processing apparatus including: a Fourier transform processing unit that converts an input signal into a frequency-domain signal for each block of size N; a multiplication processing unit that performs filter processing on the frequency-domain signal converted by the Fourier transform processing unit using filter weight coefficients; an inverse Fourier transform processing unit that performs inverse Fourier transform processing on the frequency-domain signal on which the filter processing has been performed by the multiplication processing unit; and an output signal selection unit that stores N / 2 + Δ (Δ is a positive integer) symbols centered on the output symbol at the N / 2-th position from the head among the output symbols of size N obtained by performing inverse Fourier transform processing on the inverse Fourier transform processing unit, and outputs a signal obtained by discarding output symbols other than the stored output symbols.

[0025] One aspect of the present invention is a signal processing method comprising: a Fourier transform processing step of converting an input signal into frequency domain signals in blocks of size N; a multiplication processing step of filtering the frequency domain signals converted by the Fourier transform processing step using filter weight coefficients; an inverse Fourier transform processing step of performing an inverse Fourier transform on the frequency domain signals filtered by the multiplication processing step; and an output signal selection step of saving N / 2+Δ (Δ is a positive integer) symbols centered on the N / 2th size output symbol from the beginning of the output symbols of size N obtained by performing the inverse Fourier transform processing in the inverse Fourier transform processing step, and outputting a signal in which output symbols other than the saved output symbols are discarded. [Effects of the Invention]

[0026] This invention makes it possible to reduce the amount of computation required. [Brief explanation of the drawing]

[0027] [Figure 1] This is a block diagram showing the configuration of a signal processing device. [Figure 2] This diagram shows the correspondence between the input signal sequence and the output signal sequence. [Figure 3] This is a flowchart showing the processing flow of a signal processing device. [Figure 4A] This figure shows the filter weight coefficients W and the error signal E in the conventional technology. [Figure 4B] This figure shows the filter weight coefficients W and the error signal E in this embodiment. [Figure 5A] This figure shows the overlap rate in the conventional technology. [Figure 5B] This figure shows the overlap rate in this embodiment. [Figure 6] This figure shows the reduction in computational complexity. [Figure 7] This figure shows the number of symbols per block in the case of a weakly coupled fiber. [Figure 8]This figure shows the number of symbols per block in the case of strongly coupled fibers. [Figure 9] This diagram illustrates the correspondence between the input and output signal sequences of a conventional frequency-domain MIMO-FIR signal processing system with a 50% overlap rate. [Modes for carrying out the invention]

[0028] Embodiments of the present invention will be described in detail with reference to the drawings. Figure 1 is a block diagram showing the configuration of the signal processing device 1 in an embodiment. This signal processing device 1 is a signal processing device that performs adaptive filter equalization processing in a receiving device that receives an optical signal. Note that the configuration shown in Figure 1 is a single-input single-output (SISO) configuration in order to simply explain frequency domain MIMO-FIR type signal processing. This configuration can be easily extended to a MIMO type by arranging the filter processing units 10 in parallel for the same number of input signal sequences.

[0029] Next, we will explain u(k), v(k), W(k), and d(k) shown in Figure 1. u is the time-domain input signal sequence in the kth block. v is the time-domain output signal sequence in the kth block. w is the filter weight coefficient in the kth block. d is the desired signal in the kth block. The block length is N. Regarding the oversampling rate P of the input signal, it is usually set to P=2 for optical transmission signals. However, as described in Non-Patent Literature 2, parallelizing the input signal into P sequences (i.e., setting P=1) does not lose generality, so for simplicity, we will assume P=1 in this embodiment as well. On the other hand, the number of output symbols M is M=N / 2+Δ (Δ is an integer of 1 or more less than or equal to N / 2).

[0030] The signal processing device 1 comprises a filter processing unit 10, a fast inverse Fourier transform processing unit 20, an output signal selection unit 30, an addition processing unit 40, a zero addition unit 50, and a fast Fourier transform processing unit 60. The filter processing unit 10 consists of a fast Fourier transform processing unit 11, a multiplication processing unit 12, a complex conjugate processing unit 13, a delay unit 14, a filter weight coefficient update unit 15, and multiplication processing units 16 and 17.

[0031] The Fast Fourier Transform processing unit 11 converts the input signal sequence u into a frequency domain signal U according to (1) below and outputs it to the multiplication processing unit 12 and the complex conjugate processing unit 13. In (1) below, FFT() represents the Fast Fourier Transform process. U = FFT(u) ... (1)

[0032] The complex conjugate processing unit 13 takes the complex conjugate of the input frequency domain signal U. * The output is sent to the multiplication processing unit 17. The multiplication processing unit 17 receives the output from the fast inverse Fourier transform processing unit 20, which will be described later as E(k) and U * U is obtained by taking the elemental product (Hadamard product) with * Output E to the multiplication processing unit 16. Here, the symbol "·" indicates the element-wise product operator.

[0033] The multiplication processing unit 16 controls the step size parameter μ and U * μU multiplied by E * • E is output to the filter weight coefficient update unit 15. The filter weight coefficient update unit 15 uses W(k) to calculate W+μU * The filter weight coefficients are updated by calculating E and then re-assigning the calculated value to W. The filter weight coefficient update unit 15 outputs the updated W to the delay unit 14. The delay unit 14 delays W by a predetermined timing and outputs it to the multiplication unit 12. The multiplication unit 12 outputs U·W, obtained by taking the elemental product of U and W, to the fast inverse Fourier transform unit 20.

[0034] The Fast Inverse Fourier Transform (FFT) processing unit 20 converts the signal to an output signal v according to (2) below and outputs it to the output signal selection unit 30. In (2) below, IFFT() represents the Fast Inverse Fourier Transform process. v = IFFT(U·W) ... (2)

[0035] The output signal selection unit 30 preserves the central N / 2+Δ components of v=IFFT(U·W). On the other hand, the output signal selection unit 30 discards the remaining N / 2-Δ components. The output signal sequence v output from the output signal selection unit 30 is output to the subsequent stage from the signal processing device 1 and also to the addition processing unit 40. In the conventional technology, the first N / 2 components of v=IFFT(U·W) were discarded, and the latter N / 2 components were preserved.

[0036] The addition unit 40 outputs the difference dv between the desired signal d and the output signal sequence v to the zero-adding unit 50. The zero-adding unit 50 adds zeros to dv by setting the components corresponding to the discarded output symbols to zero, and outputs it to the Fast Fourier Transform (FCR) processing unit 60. The FCR processing unit 60 corresponds to other Fourier Transform processing units. The FCR processing unit 60 calculates the error signal E in the frequency domain as shown in (3) below. The FCR processing unit 60 outputs the miscalculation signal E to the multiplication unit 17.

number

[0037] For processing at block number k+1 of the next block, W is updated by passing through the complex conjugate processing unit 13, the multiplication processing unit 17, the multiplication processing unit 16, and the filter weight coefficient update unit 15.

[0038] In the configuration shown in Figure 1 described above, as with conventional methods, an example implementation of the unconstrained frequency-domain LMS (Least Mean Square) type method (see Non-Patent Document 1) is shown as an example of the filter coefficient update algorithm. Other examples of filter coefficient update algorithms applicable to the filter processing unit include the constrained frequency-domain LMS type method and the frequency-domain RLS (Recursive Least Square) method. For the RLS method, please refer to the following literature. Zhiqun Yang, Jian Zhao, Neng Bai, Ezra Ip, Ting Wang, Zhihong Li, and Guifang Li, "Experimental demonstration of adaptive VFF-RLS-FDE for long-distance mode-division multiplexed transmission," Opt. Express 26, 18362-18367 (2018)

[0039] Figure 2 shows the correspondence between the input signal sequence and the output signal sequence of the frequency-domain MIMO-FIR type signal processing described above. In Figure 2, the k-th block and the (k+1)-th block are shown as the input signal sequence. The output signal sequence is also shown corresponding to the input signal sequence.

[0040] For the k-th block shown in Figure 2, the output signal is N / 2+Δ symbols, which are the central N / 2 symbols centered around the N / 2-size output symbol from the beginning, plus the Δ symbol. It is possible to reduce the overlap rate by the amount of this increment of Δ symbols, and the overlap rate in this case is given by (4) below. Overlap rate = Δ / N…(4)

[0041] In other words, since Δ is an integer less than or equal to N / 2, the input signal sequence of the (k+1)th block can be configured in a way that reduces the overlap rate to 50%. Note that the "signal processing scale per block (complex multiplication number)" is the same for both the conventional method explained in Figure 9 and the method shown in this embodiment explained in Figure 1. On the other hand, the number of output symbols per block increases in this embodiment, so according to this embodiment, the "signal processing scale per output symbol" can be effectively reduced.

[0042] Figure 3 is a flowchart showing the processing flow of the signal processing device 1 described above. The fast inverse Fourier transform processing unit 11 converts the input signal into frequency domain signals in blocks of size N (step S101). The multiplication processing unit 12 and the Fourier transform processing unit 11 perform filtering on the frequency domain signals converted by the Fourier transform processing unit 11 using filter weight coefficients (step S102). The fast inverse Fourier transform processing unit 20 performs inverse Fourier transform on the frequency domain signals that have been filtered by the multiplication processing unit 12 (step S103). The output signal selection unit 30 saves N / 2 + Δ (Δ is a positive integer) symbols from the output symbols of size N obtained by the inverse Fourier transform processing performed by the fast inverse Fourier transform processing unit 20, centering on the N / 2 size output symbol from the beginning, and outputs a signal with output symbols other than those saved discarded (step S104).

[0043] Next, we will explain the embodiment described above in comparison with the prior art. First, Figure 4A shows the filter weight coefficients W and the error signal E in the prior art. As shown in Figure 4A, W is obtained by performing a Fast Fourier Transform on N / 2 w and a zero vector containing N / 2 zeros. E is obtained by performing a Fast Fourier Transform on a zero vector containing N / 2 zeros and the difference e between the desired signal d and the output signal sequence v.

[0044] Figure 4B shows the filter weight coefficients W and the error signal E in this embodiment. As shown in Figure 4B, W is obtained by performing a Fast Fourier Transform on N / 2 w, a zero vector containing N / 2 zeros, and N / 2 w. E is obtained by performing a Fast Fourier Transform on a zero vector containing N / 4 zeros, the difference e between the desired signal d and the output signal sequence v, and a zero vector containing N / 4 zeros. Compared to the prior art, in this embodiment the algorithm has been modified so that the output symbol, which is not subject to inter-block interference, is placed in the center.

[0045] Next, the overlap rate will be explained in comparison between this embodiment and the prior art. Figure 5A shows the overlap rate in the prior art. Figure 5A shows the k-th block and the (k+1)-th block. The output signal sequence is shown corresponding to the input signal sequence. As shown in Figure 5A, the overlap rate in the prior art was 50%.

[0046] Figure 5B shows the overlap rate in this embodiment. Figure 5B shows the k-th block and the (k+1)-th block. The output signal sequence is shown corresponding to the input signal sequence. As shown in Figure 5B, the overlap rate can be reduced to a range where there is no effective interference within the block (up to a maximum of 25%).

[0047] Next, we will explain the reduction in computational complexity compared to conventional technology. In the following explanation, the conventional frequency-domain MIMO-FIR signal processing with an overlap rate of 50% will be referred to as Method A. The frequency-domain MIMO-FIR signal processing with a reduced overlap rate according to this embodiment will be referred to as Method B.

[0048] Figure 6 shows the computational complexity reduction rate. In the graph shown in Figure 6, the horizontal axis represents distance, and the vertical axis represents the computational complexity reduction rate. The computational complexity reduction rate is defined as the signal amount required for method B divided by the signal processing scale of method B. In the computational complexity reduction rate shown in Figure 6, the complex multiplier number is used as a parameter representing the computational complexity. The complex multiplier number required for the fast (inverse) Fourier transform processing of a signal with block length N (where N is a power of 2) was set to N / 2 × log(N) (base of the logarithm is 2). Furthermore, a coupled multicore fiber with 12 cores was assumed as the type of optical fiber constituting the optical transmission path. The symbol rate of the signal was set to 10 GBaud. The oversampling rate was set to 2. The spatial mode dispersion coefficient was set to 20 ps / (km)^1 / 2. Note that the symbol "^" indicates exponentiation, in this case representing the square root.

[0049] In Figure 6, a triangular wave-shaped graph of the computational complexity reduction rate is repeatedly shown. This transition represents a point where the block length, based on the Fast Fourier Transform processing, transitions in powers of 2. The value of Δ is set to the condition that no interference occurs within the block. As shown in Figure 6, all points except those located on the triangular wave are less than 1. This indicates that the computational complexity is reduced by this embodiment. The average computational complexity reduction rate between distances of 1 km and 3000 km shown in Figure 6 was 0.82. Thus, according to this embodiment, the computational complexity can be reduced compared to the conventional technology.

[0050] Next, we will explain the number of symbols per block compared to the conventional technique in the case of unconstrained FDE (frequency domain equalization)-LMS (least-mean square). Figure 7 shows the number of symbols per block in the case of weakly coupled fibers. Figure 8 shows the number of symbols per block in the case of strongly coupled fibers. In Figures 7 and 8, the horizontal axis represents distance, and the vertical axis represents the number of symbols per block. The solid line shows the number of symbols per block when this embodiment is applied, and the dashed line shows the number of symbols per block when the conventional technique is applied.

[0051] The calculation conditions for the number of symbols per block shown in Figure 7 are an SMD coefficient of 20 ps / km, a symbol rate of 10 GBaud, 12 modes, and a scaling factor of 1. The calculation conditions for the number of symbols per block shown in Figure 8 are an SMD coefficient of 20 ps / km, a symbol rate of 10 GBaud, 12 modes, and a scaling factor of 6.

[0052] As shown in Figures 7 and 8, it can be seen that the number of symbols per block in this embodiment is higher than that of the prior art at almost every distance.

[0053] While embodiments of this invention have been described in detail above with reference to the drawings, the specific configuration is not limited to these embodiments and includes designs and the like that do not depart from the spirit of this invention. [Industrial applicability]

[0054] The present invention is applicable to a receiving device that receives optical signals. [Explanation of Symbols]

[0055] 1...Signal processing unit, 10...Filter processing unit, 11...Fast Fourier Transform processing unit, 12...Multiplication processing unit, 13...Complex conjugate processing unit, 14...Delay unit, 15...Filter weight coefficient update unit, 16...Multiplication processing unit, 17...Multiplication processing unit, 20...Fast Inverse Fourier Transform processing unit, 30...Output signal selection unit, 40...Addition processing unit, 50...Zero addition unit, 60...Fast Fourier Transform processing unit

Claims

1. A Fourier transform processing unit that converts the input signal into frequency domain signals in blocks of size N, A multiplication processing unit performs filtering on the frequency domain signal transformed by the Fourier transform processing unit using filter weight coefficients, An inverse Fourier transform processing unit performs an inverse Fourier transform on a frequency domain signal that has been filtered by the multiplication processing unit, An output signal selection unit that, from the output symbols of size N obtained by the inverse Fourier transform process, saves N / 2 + Δ (Δ is a positive integer) symbols centered on the N / 2th size output symbol from the beginning, and outputs a signal in which output symbols other than those saved are discarded. An update unit updates the filter weight coefficients so that the output symbol, which is not subject to interfering within the block, is placed in the center of the frequency domain signal. A signal processing device equipped with a signal processing device.

2. The system includes a zero-adding unit that outputs a signal obtained by adding zeros to a difference vector formed from the difference between the signal output by the output signal selection unit and the desired signal. The signal processing apparatus according to claim 1, wherein the zero-adding unit sets the component corresponding to the output symbol discarded by the output signal selection unit to zero.

3. The signal processing apparatus according to claim 2, further comprising another Fourier transform processing unit that outputs a signal obtained by Fourier transforming a signal to which zeros have been added by the zero-adding unit.

4. A complex conjugate processing unit performs complex conjugate processing on the signal output by the Fourier transform processing unit, Equipped with, The signal processing apparatus according to claim 3, wherein the update unit updates the filter weight coefficients using the signal that has been complex conjugated by the complex conjugate processing unit and the signal output by the other Fourier transform processing unit.

5. A Fourier transform processing step in which the Fourier transform processing unit converts the input signal into frequency domain signals in blocks of size N, A multiplication processing unit performs a multiplication processing step in which the frequency domain signal transformed by the Fourier transform processing step is filtered using filter weight coefficients, An inverse Fourier transform processing step in which the inverse Fourier transform processing unit performs an inverse Fourier transform on the frequency domain signal that has been filtered by the multiplication processing step, The output signal selection step involves the output signal selection unit saving N / 2 + Δ (Δ is a positive integer) symbols centered on the N / 2th size output symbol from the output symbols of size N obtained by the inverse Fourier transform process, and outputting a signal in which all output symbols other than those saved are discarded. The update step involves updating the filter weight coefficients so that the output symbol, which is not subject to block interference, is centered in the frequency domain signal, by the update unit. A signal processing method equipped with [a specific feature].

Images