Research Of Digital FIR Filters Based On SOPC Technology And HPSO Algorithm

As the digital signal processing technology has been widely used in aerospace, geological survey, medical engineering, digital TV, power system and other fields, digital filter is an indispensable component of the hardware products in the above fields. FIR filter system has many advantages, such as stable performance, strict linear phase and arbitrary amplitude. Its application research is of great theoretical significance and practical value. The implementation process mainly includes the selection of filter coefficients and optimization of the structures for convolution operation.The process of selecting filter coefficients was converted to the multi-parameter optimization problem. The main strategy of improved particle swarm optimization algorithm, such as optimal learning factor and inertia weight, fuse with other intelligent algorithms and adjustment of the related parameters were discussed. In stead of the PSO method for updating the position of particle information by tracking the extremum trajectory, HPSO(Hybrid Particle Swarm Optimization) which adopts the natural selection mechanism and the cloud for dynamically adjusting inertia weight to constraint the flight path of the particle was introduced in the thesis, so as to coordinate the ability of the algorithm on local development and the global exploration. In order to prevent the convergence of emergence at the later evolution and the stagnation phenomenon of each particles, Gauss disturbance was adopted to search the global optimal solution. The whole particle swarm is divided into multiple subgroups. Subgroup and elite particles search step by step with collaborative evolving, thus each particle not only keeps the optimization direction of the subgroups, but also be closed to the direction of global optimal particles.Substantially, FIR convolution is the mathematical operations of multiplying and accumulating between the input data and the filter coefficients. According the direct structure which can be easily realized in FPGA, the serial, parallel and distributed structures used in convolution operation were analyzed. SOPC taken as the core technology constructs, Nios II processor system were built in the FPGA chip under Qsys environment. Convolution, ADC and DAC kernels were customed using HDL language. Modules of time delay reset and pll were designed. Tcl script file was established. Thus, the FPGA top-level hardware circuit was ultimately completed. The design of software for soft-core processors was completed through the NiosII SBT under Eclipse environment based on a BSP file.The simulation results show that FIR filters optimized by HPSO algorithm have the advantages of small fluctuation in passband and big attenuation in stopband, which verifies the improved performances of the optimized FIR filters. A low pass filter taked as an example, testbench simulation test files were writen using ModelSim simulation software. Results were displayed by the MATLAB software, which shows that the mixed high-frequency sinusoidal noise signal has been filtered, meeting the requirements of preliminary design, so the validity and feasibility of these methods are proved.