Efficient Implementations of Complex and Real Valued Filter Banksfor Comparative Subband Processing with an Application to

Abstract. In this paper, we discuss efficient methods to implement correlation type algorithms in subbands.Based on a polyphase representation, a modulated GDFT filter bank with arbitrary integer decimation ratio can be performed at low cost, yielding complex subband signals. Real valued subbands can be achievedby appropriate postprocessing of the complex filter bank. Together with consideration of the computational order of the algorithm applied for subband processing, complex or real implementation may be better suited,for which we present criteria. For the example of subband adaptive filtering as applied in acoustic echo cancellation, we compare a number of algorithms in terms of the most efficient implementation. 1. Introduction Often the processing of signals is performed in frequency bands. This signal decomposition allows a lower-ing the sampling rate in the subbands, thus reducing the computational complexity of implemented systems, and leads to a convenient parallelization. For comparing the subbands of different signals as e.g. performedin adaptive filtering, critical decimation by a facto