Multiple window non-linear time-varying spectral analysis

A non-linear multi-window method for generating a time-varying spectrum of non-stationary signals in noise is presented. The timevarying spectrum is computed from an optimally weighted average of multiple Hermite windowed spectrograms. The weights are determined using linear least squares estimation with respect to a reference time-frequency distribution. A masking operation is also used to reduce extraneous side lobes introduced by higher order Hermite windows. Several examples are provided, with performance criteria measures, to demonstrate and quantify the effectiveness of this new method.