Discrete prolate spheroidal sequence based filter banks for the analysis of nonstationary signals

Abstract

Filter banks are often used in the analysis and the synthesis of signals. By using cascade or parallel connected filters, multi-resolution analysis of a signal in different sub-bands is possible. A method based on the filter banks using discrete prolate spheroidal sequences (DPSS) is proposed in this paper. DPSS are solutions to an energy maximization problem in a limited bandwidth context; their time-frequency concentration aspect is exploited in the proposed design. A filter bank is derived using DPSS sequences and their modulated counter parts and is utilized in the analysis of non-stationary signals. The designed filter-bank achieves a concise representation of a non-stationary signal by maximizing the spectral energy in each sub-band.