Using resonating filter banks and energy levels to detect signal transitions in complex sound fields

In the quest to develop sound analysis tools much work has been put into time-frequency analysis techniques. Here we summarize an effort to utilize filterbanks composed of digital resonators as a time-frequency analysis tool. The resonators are designed to have constant Q and as such share some properties with the popular wavelet analysis techniques. The resonator output is coupled with an energy analysis technique. These computed energy levels are used to detect transitions in individual bands of the filterbank for the purposes of sound source extraction from a complex sound signal.