一种基于遗传算法的混合滤波器组设计方法

针对可满足近似完全重构的双通道混合滤波器组,其中高阶数的模拟滤波器一般不容易设计优化。采用遗传算法设计5阶模拟分解滤波器,并基于逆快速傅里叶变换实现数字综合滤波器的设计优化以滤除掉镜像频谱,保证近似完全重构。文中设计了由5阶模拟分解滤波器和32阶数字综合滤波器组成的混合滤波器组,仿真结果表明:可以实现的最大失真误差为4.761 8×10-11dB,平均失真误差为-9.2×10-14dB,最大混叠误差为-154 dB,平均混叠误差为-200 dB,可满足24 bits的模数转换器系统的要求。

A Method of Hybrid Filter Banks Design Based on Genetic Algorithm

High-order analog filter is difficult to design and optimize for designing two-channel hybrid filter banks which meet the need of near of perfect reconstruction. Genetic algorithm was used in this paper to design 5-order analog analysis filter. Moreover, the design and optimization of digital synthesis filters in order to filter the mirror spectrum was realized by it which ensure near perfect reconstruction based on IFFT. Hybrid filter banks which consist of 5-order analog analysis filters and 32-order digital synthesis filters were designed in this paper. Maximum distortion error of 4.761 8×10-11 dB, average distortion error of -9.2×10-14 dB, maximum aliasing error of -150 dB and average error of -200 dB can be achieved in the hybrid filter banks according to the simulation results. The newly designed hybrid filter banks to meet the requirement of 24 bits analog-to-digital converter systems well is concluded.