Aliasing-free criticalsampled filter banks design based on frequency domain decomposition

In subband adaptive signal processing, the performance of system is seriously disturbed by the aliasing in-band introduced by decimator, which comes from the none-ideal of analysis filter banks. In this paper a novel method called DFTSD (discrete Fourier transform subband decomposition) is proposed for signal decomposition in frequency domain, in which different subband signals are acquired by line spectrum abstraction. In this proposed method the aliasing is eliminated radically by using the equivalent analysis filter bank with ideal frequency response. The conditions of perfect reconstruction (PR) as well as the important issues of design are analyzed. The experiments show that the signals decomposition/reconstruction can be realized perfectly with DFTSD and the average error of reconstruction is nearly zero.