Automatic quality assessment and optimisation of quadratic time–frequency representations

A fully automatic method for optimising the smoothed-pseudo Wigner-Ville distribution is presented. An alternative view on time-frequency representations (TFRs) through correlation intensity landscapes is offered, and the notion of wanted and unwanted correlations and their balance is introduced which is introduced as the measure of quality. The problem of optimising TFRs is then formulated as the cost function based on this measure. With the differential evolution algorithm as the solver, the method´s performance is tested on synthetic and real-life signals.