An automatic fast optimization of Quadratic Time-frequency Distribution using the hybrid genetic algorithm

This paper presents a novel framework for a fully automatic optimization of Quadratic Time-frequency Distributions (QTFDs). This 'black box' approach automatically adjusts the QTFD kernel parameters by using a hybrid genetic algorithm (HGA). This results in an optimal use of QTFDs suitable for non-specialist users without requiring any additional input except for the signal itself. This optimization problem has been formulated as the minimization of the cost function of a modified energy concentration measure. The efficiency of the proposed method has been demonstrated by representing selected non-stationary signals in the time-frequency domain and testing robustness under different SNR conditions by estimating the instantaneous frequency. A fast implementation of QTFD optimization reduces computation time significantly; e.g., the computation time of a real world bat signal of 400 samples reduces to 3.5885ź0.3942s from its standard implementation (53.0910ź1.445s). Optimized selection of parameters for Quadratic Time-Frequency Distributions (QTFDs).A novel hybrid genetic algorithm is proposed for a full optimization of QTFDs.Offers a 'black box' approach that needs no extra input except the signal itself.The fast and memory efficient implementation is advantageous in 'big data' science.Useful for the non-specialist users to optimally use QTFDs in many disciplines.