Time/frequency techniques for signal feature detection

A method for detecting linear chirps in audio signals is presented. It uses a complex uniform spaced filter bank as a basis for the detection. The filter bank has an efficient implementation, good frequency localization, and delivers magnitude and phase of the subband signals. The complex filter bank uses two parallel real valued filter banks, each with critical sampling and perfect reconstruction capability. Detection is accomplished via a template based thresholding scheme. Various templates are used to account for differing amounts of linear chirp in the data sets. Experimental results using actual audio data are presented, which demonstrate the method is a useful tool for the detection of linear chirps in audio signals.