The Implementation And Application Of Fast Filter Bank Design Based On Low Complexity Extrapolated Impulse Response Method

In this paper,the issue of high complexity of digital filter banks in digital communication module is discussed.We focus on the low complexity of the extrapolated impulse response method of fast filter bank,and study the design of EIR method,the tree structure path coding of FFB,the prototype filter design,the equivalent structure derivation and the multi-phase structure.The low complexity design of EIR-FFB is studied,and a uniform and non-uniform digital filter bank is designed and applied to digital channelization.The main work of this paper is as follows:Firstly,the prototype EIR method is studied.Due to the sidelobe expansion is performed by quasi-periodic between the sidelobes of the FIR filter impulse response,the design complexity is reduced.Satisfied initial coefficients are obtained by coefficient-autocorrelation-matrix decomposition method.On the basis of this,the main lobe coefficient,the sidelobe coefficient and the extrapolation ratio factor form the coefficient vector to be optimized.The second-order cone programming is used to find the optimal solution in the frequency domain of the half-band filter.Secondly,two equivalent structures of FFB are deduced.According to the given global design parameters,the design flow of the prototype filter is given,and the overall design of FFB is realized.The relationship between the sub-filters and the channel response is determined by the path coding of the tree structure,and then two equivalent structures of FFB are deduced:uniform filter bank and polyphase filter bank.Due to the frequency shift characteristics between the channel responses,FFB is equivalent to the uniform filter bank.The general form of polyphase structure is deduced.Besides,by the channelized receiver structure with complex input signal,the polyphase equivalent implementation of FFB is given.Subsequently,the low complexity design of EIR-FFB is proposed,further reduces the design complexity of the FFB.Four classic filter banks are analyzed:Discrete Fourier Transform Filter Bank(DFTFB),Cosine Modulated Filter Bank(CMFB),Modulated Perfect Reconstruction Filter Bank(MPRB),Multi-Resolution Filter Bank(MRFB).The comparison between EIR filter and prototype filter,EIR-FFB and prototype FFB is given by simulation.Compared to the above four filter banks,EIR-FFB has lower complexity under the same design parameters.Finally,the design of EIR-FFB uniform and non-uniform digital channelization is realized.The Coefficient Decimation Method(CDM)and Modified Coefficient Decomation Method(MCDM)are analysed through four implementations:CDM-?,CDM-?,MCDM-? and MCDM-?.An example of simulating a uniform and non-uniform digital filter bank using Improved Coefficient Decomation Method(ICDM)is presented.Under the same design parameters,the EIR-FFB channelization design steps are given,and the EIR-FFB uniform and non-uniform digital channelization is realized.The simulation results show that the proposed EIR-FFB has excellent performance and low complexity.