An adaptive optimal-kernel time-frequency representation

Signal-dependent time-frequency representations perform well for a much wider range of signals than any fixed-kernel distribution. The time-frequency representation presented here, based on a signal-dependent radially Gaussian kernel that adapts over time, tracks signal component variations over time and supports online implementation for signals of arbitrary length. The method uses a short-time ambiguity function for kernel optimization and as an intermediate step in computing constant-time slices of the time-frequency representation. While somewhat more expensive than fixed-kernel representation, this technique often provides much better performance.<>