A fast signal-dependent time-frequency representation

In last few years, in order to overcome same limitations of the short time Fourier transform (STFT), while avoiding the cross-terms that make the Wigner distribution difficult to interpret, some signal-dependent time-frequency representations (SDTFR) have been proposed. In this paper, the authors introduce a computationally efficient signal-dependent time-frequency method which is suitable for on-line analysis. This SDTFR uses a Gaussian window (GW) similar to STFT, but varies the parameter σ of the GW with time to achieve high signal concentration and high resolution in time. The parameter σ can be automatically calculated by the slope of instantaneous frequency (IF) and instantaneous bandwidth (IB) at that time