SRMD: Sparse Random Mode Decomposition

Time-frequency analysis is an important tool for analyzing and leveraging information from signals. Various time-frequency approaches lead to a particular decomposition of signals into important time-varying features which can illuminate intrinsic behaviors, be used for analytics, or assist in smoothing the signal. For example, the classical short-time Fourier transform (STFT) extracts a time-varying representation by localizing the signal in time using a finite window length and applying the Fourier transform to each localized segment. Since the window length is fixed, the resulting analysis leads to a uniform time-frequency resolution. The continuous wavelet transform (CWT) constructs a representation of the signal using a variable time window by scaling the mother wavelet and thus leads to a multiscale representation of the signal.