Low-complexity segmented filtering method based on bilateral overlap-save method in broadband communication system

Abstract

The invention belongs to the technical field of digital signal processing, and particularly relates to a low-complexity segmented filtering method based on a bilateral overlap-save method in a broadband communication system. Aiming at the problem of realizing a large-scale finite impulse response filter in a hardware platform with a limited Fourier transform size, the method comprises the following steps of: dividing a long filter into a plurality of sub-segments and respectively performing frequency domain representation; the method comprises the following steps: performing bilateral overlapping storage on input signal blocks, performing fast Fourier transform, introducing phase shift in a frequency domain to realize time shift, further accumulating frequency domain product results of each filter section and a signal block, and finally obtaining filtering output only through one-time inverse fast Fourier transform. According to the method, the inverse fast Fourier transform operation times are remarkably reduced, the calculation complexity and the storage requirement are greatly reduced, the filtering performance is kept, meanwhile, excellent hardware friendliness is achieved, and the method is particularly suitable for occasions needing an ultra-long filter, such as optical fiber communication dispersion compensation processing.

Description

Technical Field

[0001] This invention relates to the field of digital signal processing, and more specifically to a low-complexity piecewise filtering method applicable to the implementation of large-scale finite impulse response (FIR) filters, particularly a low-complexity piecewise filtering method based on the bilateral overlap-preservation method in broadband communication systems. Background Technology

[0002] In modern communication systems, especially long-distance fiber optic transmission, large-scale finite impulse response (FIR) filters with lengths exceeding 10,000 taps are often required to compensate for channel impairments such as dispersion. Traditional overlap-save (OLS) or overlap-add (OLA) methods significantly reduce computational complexity by dividing the signal into blocks and convolving them in the frequency domain, but require the Fast Fourier Transform (FFT) size B to be larger than the filter length F. In hardware-constrained scenarios (such as FPGAs and ASICs), the FFT size is often much smaller than the filter length, and directly applying OLS can lead to performance degradation or make it impossible to implement.

[0003] An intuitive improvement is to segment the filter (VOSF method), with each segment shorter than the FFT size, and perform OLS on each segment separately. However, this method requires performing FFT and Inverse Fast Fourier Transform (IFFT) separately for each filter segment, leading to a sharp increase in computational and storage requirements and inefficiency.

[0004] Therefore, there is an urgent need for a low-complexity piecewise filtering method that can still efficiently implement large-scale filters under the condition of FFT size constraints. Summary of the Invention

[0005] To address the shortcomings of the existing technologies, the present invention aims to propose a low-complexity segmented filtering method (OLS-BS-SF) based on the bilateral overlap-preservation method in broadband communication systems. By implementing time shifting in the frequency domain, the convolution results of multiple filter segments are merged into a single IFFT, thereby significantly reducing the number of IFFT operations and achieving a substantial reduction in complexity. The present invention can efficiently implement large-scale filters under hardware conditions with limited Fast Fourier Transform (FFT) size, and is suitable for dispersion compensation in optical fiber communication systems, baseband processing in wireless communication, and other scenarios requiring long filters.

[0006] The technical solution of the present invention is described in detail below.

[0007] This invention provides a low-complexity piecewise filtering method based on the bilateral overlap-preservation method in broadband communication systems, comprising the following steps:

[0008] (1) Signal and filter segmentation

[0009] Input signals are divided into block lengths Block division, overlap length of adjacent signal blocks ; The large-scale filter of length F is divided into lengths Divided into Each filter segment is zero-padded to its length. And perform a Fast Fourier Transform (FFT) to obtain the frequency domain coefficients. ;

[0010] (2) Frequency domain shifting and convolution

[0011] Perform a Fast Fourier Transform (FFT) once on each signal block to obtain the b-th frequency domain sub-signal block. Introducing phase shift in the frequency domain and utilizing frequency domain phase factors To achieve time shifting, the shifted bm-th frequency domain sub-signal block With the corresponding frequency domain coefficients Multiply;

[0012] (3) Frequency domain aggregation and single IFFT

[0013] Add the frequency domain products of all filter segments to obtain the aggregated frequency domain result; perform an Inverse Fast Fourier Transform (IFFT) on the aggregated result, discarding the unaggregated values. Only the valid portion of the sampling points is retained.

[0014] (4) Output splicing

[0015] The effective outputs of each block are concatenated in sequence to obtain the final filtering result.

[0016] In this invention, the filter segment length With signal block length The relationship is It can guarantee the frequency domain

[0017] Shifting is equivalent to block shifting, thus enabling single IFFT sharing. This reduces the number of IFFT operations to only once per signal block, thereby lowering complexity.

[0018] In this invention, in step (2), the frequency domain shift is achieved by adjusting the input signal block index. With filter segment index The correspondence is achieved by shifting the input signal block forward. Spacing between blocks , .

[0019] In this invention, when the method is applied to dispersion compensation in an optical fiber communication system, it achieves efficient filtering under FFT size constraints. The filter is a band-limited dispersion-compensated CDC filter. The FFT size is 512 or 1024 points.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0021] (1) Reduced FFT / IFFT times: Each signal block only needs one FFT and one IFFT.

[0022] (2) Hardware friendly: suitable for FFT size-constrained implementation platforms.

[0023] (3) Performance Preservation: Even when the filter length is much larger than the FFT size, it can still maintain performance comparable to the full-length OLS.

[0024] Filtering performance.

[0025] (4) Wide range of applications: It can be used for fiber dispersion compensation (CDC), wireless communication baseband filtering, radar signal processing, etc. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of a low-complexity piecewise filtering method based on the bilateral overlap-preservation approach. Detailed Implementation

[0027] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0028] The schematic diagram of a low-complexity piecewise filtering method based on the bilateral overlap-preservation approach is shown below. Figure 1 As shown, this invention divides the original signal of length D into overlapping sub-signal blocks of length B, where OL is the overlap length, and then transforms them to the frequency domain using FFT; a long filter divides the signal in the time domain and transforms it to H using FFT. i (i=0,1,2…M-1), and multiply with the signal in the frequency domain. The frequency domain products of all filter segments are summed to obtain the aggregated frequency domain result. An IFFT is performed on the aggregated result, discarding the first OL samples and retaining only the valid parts. Finally, the valid outputs of each block are concatenated in sequence to obtain the final filtered result of length D.

[0029] This invention provides a low-complexity segmented filtering method based on the bilateral overlap-preservation method in broadband communication systems. The method includes: dividing a long filter into multiple sub-segments and representing each sub-segment in the frequency domain; performing bilateral overlap-preservation on the input signal blocks and executing a Fast Fourier Transform (FFT); introducing a phase shift in the frequency domain to achieve time shifting; accumulating the frequency domain product results of each filter segment and the signal block; and finally obtaining the filtered output through only one Inverse Fast Fourier Transform (IFT). This invention significantly reduces the number of IFT operations, greatly reducing computational complexity and storage requirements. It is particularly suitable for applications requiring ultra-long filters, such as dispersion compensation in fiber optic communication, and maintains excellent filtering performance while possessing excellent hardware friendliness.

[0030] The method described in this invention specifically includes four parts: "signal and filter segmentation", "frequency domain shifting and convolution", "frequency domain aggregation and single IFFT", and "output splicing".

[0031] (1) Signal and filter segmentation

[0032] Input signals are divided into block lengths Block division, overlap length of adjacent blocks (50% overlap). Large-scale filters of length F are arranged according to length... Divided into Each sub-segment is zero-padded to its length. And perform FFT to obtain frequency domain coefficients. .

[0033] (2) Frequency domain shifting and convolution

[0034] Perform an FFT once for each signal block to obtain Using frequency domain phase factors To achieve time shifting, the shifted time... With the corresponding Multiply.

[0035] (3) Frequency domain aggregation and single IFFT

[0036] The frequency domain products of all filter segments are summed to obtain the aggregated frequency domain result. An IFFT is then performed on the aggregated result, discarding the unaggregated segments. Only the valid portion of the sampling points is retained.

[0037] (4) Output splicing

[0038] The effective outputs of each block are concatenated in sequence to obtain the final filtering result.

[0039] Application Example – Fiber Dispersion Compensation (CDC)

[0040] In fiber optic communication systems, the frequency response of a CDC filter is:

[0041]

[0042] in, For group velocity dispersion parameters, This represents the length of the optical fiber.

[0043] The method of this invention can efficiently implement CDC filtering under conditions where the FFT size is limited (such as 512 or 1024 points). The results show that, with a bandwidth of 200 GHz, an FFT size of 1024 points, a large-scale filter length F of 16384 points, and a signal block length B of 1024 points, this invention can reduce the computational complexity by about 94% compared with the traditional segmented OLS (VOSF).

[0044] In summary, this invention proposes an improved bilateral overlap-preservation segmented filtering algorithm. This method achieves efficient implementation of large-scale filters under hardware constraints limited by FFT size, and significantly reduces the number of inverse fast Fourier transform operations (only once per signal block), greatly reducing computational complexity and storage requirements. It is particularly suitable for applications requiring ultra-long filters, such as dispersion compensation in fiber optic communication, and possesses excellent hardware friendliness while maintaining filtering performance.

Claims

1. A low-complexity piecewise filtering method based on the bilateral overlap-preservation method in a broadband communication system, characterized in that, Includes the following steps: (1) Signal and filter segmentation Input signals are divided into block lengths Block division, overlap length of adjacent signal blocks ; Distribute large-scale filters of length F according to length Divided into Each filter segment is zero-padded to its length. And perform a Fast Fourier Transform (FFT) to obtain the frequency domain coefficients. ; (2) Frequency domain shifting and convolution Perform a Fast Fourier Transform (FFT) once on each signal block to obtain the b-th frequency domain sub-signal block. Introducing phase shift in the frequency domain and utilizing frequency domain phase factors To achieve time shifting, the shifted bm-th frequency domain sub-signal block With the corresponding frequency domain coefficients Multiply; (3) Frequency domain aggregation and single IFFT Add the frequency domain products of all filter segments to obtain the aggregated frequency domain result; perform an Inverse Fast Fourier Transform (IFFT) on the aggregated result, discarding the unaggregated values. Only the valid portion of the sampling points is retained. (4) Output splicing The effective outputs of each block are concatenated in sequence to obtain the final filtering result.

2. The low-complexity piecewise filtering method according to claim 1, characterized in that, In step (2), the frequency domain shift is achieved by adjusting the input signal block index. With filter segment index The correspondence is achieved by shifting the input signal block forward. Spacing between blocks , .

3. The low-complexity piecewise filtering method according to claim 1, characterized in that, When the method is applied to dispersion compensation in optical fiber communication systems, it achieves efficient filtering under FFT size constraints. The filter is a band-limited dispersion-compensated CDC filter.

4. The low-complexity piecewise filtering method according to claim 3, characterized in that, The FFT size is 512 or 1024 points.

Images