Evolutionary spectrum estimation by positivity constrained deconvolution

A deconvolution technique to estimate the evolutionary spectrum (ES) of nonstationary signals by deconvolving the blurring effects of the kernel function from bilinear time-frequency distributions (TFD) is presented. The resulting ES has desirable properties such as positivity, higher resolution, higher concentration in time-frequency. The proposed algorithm is computationally more efficient compared to the recently proposed entropy based deconvolution method. Unlike the entropy method the new algorithm can be adapted to deconvolve TFDs other than the spectrogram