基于小波包子带分解回声对消移不变性的研究

A STUDY ON SHIFTABILITY OF ECHO CANCELLATION BASED ON WAVELET SUBBANDS DECOMPOSITION

在自适应回声对消中应用多分辨率分析———对信号进行小波包子带分解后实行自适应对消，可以极大地降低计算复杂性．但若对信号进行的是正交分解（用Ｄａｕｂｅｃｈｉｅｓ小波），过程不具移不变性．为使过程具近似移不变性，作者由Ｄａｕｂｅｃｈｉｅｓ小波基（Ｄ２０）开始，通过修正尺度函数，再由双尺度方程得到小波基，使小波基近似移不变，同时还考虑了修正后滤波器的完美重构问题，并讨论了移不变性和完美重构之间的关系．

The application of multiresolution analysis (MARS) in adaptive echo cancellation—adaptive echo cancellation following the signal’s wavelet packet subband decompsition,can reduce the computational complexity greatly.But,if an orthogonal decomposition is used,the decomposition process will have no shiftability.To make the process have shiftability,we reshape a starting scaling function(the Daubechies D20 scaling function),then obtained the new wavelet filter coeffitions using the 2 scale equation,thus the wavelet filter is nearly shiftable.The authors also considered the filter perfect reconstruction(PR) after the filter was reshaped and discussed the relationship between filter’s shiftability and perfect reconstruction.