基于功率互补对的树形结构混合滤波器组

提出了一种近似完全重构的树形结构混合滤波器组(HFB)的设计方法,利用基于功率互补对的树形结构的模拟分解滤波器,而非传统的低通、带通和高通分解滤波器均匀分割整个系统带宽.采用遗传算法实现了功率互补对的优化设计,并基于逆快速傅立叶变换(IFFT)实现了数字综合滤波器的设计以保证近似完全重构.3阶和5阶的功率互补对与63阶的数字FIR(有限冲击响应)综合滤波器所组成的四通道混合滤波器组,具有1.06×10^-7dB平均失真误差和-111?dB的平均混叠误差,满足15?bit的模数转换(ADC)系统要求.

Tree structured hybrid filter banks based on power complementary pairs

A near-perfect reconstruction tree-structured hybrid filter bank (HFB) was presented. The analog analysis filters to analyze input signals are not the low-pass, band-pass or high-pass filters designed by using conventional methods, but the tree-structured analog filters based on power complementary pairs,whose parameters were optimized by genetic algorithm. And the digital synthesis filters were implemented by means of the optimization method based on inverse fast Fourier transforms. 3rd-order and 5th-order analog power complementary filter pairs were designed for 4-channel tree-structure HFB. And length 63 FIR(finite impulse respouse) synthesis filters were implemented to ensure the near-perfect reconstruction. The HFB, which yields 1.06×10^-7dB average deviation from 0dB distortion and -111dB average aliasing, is suitable for systems up to 15 bits resolution.