Iterative Synchrosqueezing-Based General Linear Chirplet Transform for Time-Frequency Feature Extraction
Yi Liu, Hang Xiang, Zhansi Jiang, Jiawei Xiang
Abstract
Short-time Fourier transform (STFT)-based methods are widely applied in industrial areas. However, these methods are inadequate to process non-stationary signals under variable operating conditions. An improved general linear chirplet transform method is developed by iteratively upgrading the instantaneous frequency (IF) and introducing a synchrosqueezing operator simultaneously. Initially, an iterative upgrading strategy is adopted to improve the estimation accuracy of the IF curves. Then, a synchrosqueezing operator is employed to enhance the concentration of the time-frequency representation under variable operating conditions. Finally, experiments that utilize simulated data are conducted to verify the effectiveness. Experimental results show that the enhanced time-frequency analysis (TFA) method can sharpen IF curves and enhance the time-frequency readability compared with other advanced TFA methods. Moreover, the feature extraction ability of the present method is superior to other commonly used methods.