Stochastic time–frequency-entropy feature extraction method

Entropy-based feature extraction methods have shown marked effectiveness across various fields. However, entropy measures are susceptible to noise interference and typically consider only time-domain information, which leads to suboptimal feature extraction performance. To address these limitations, this paper proposes a novel method: stochastic time–frequency entropy feature extraction. First, the signal-to-noise ratio of the input signal is enhanced through stochastic resonance; then, the time–frequency representation is obtained via wavelet transforms and converted into multivariate signals. Ultimately, the complexity of the multivariate signal is calculated via multivariate slope entropy. Validated through simulation and experimental data analysis, stochastic time–frequency entropy demonstrates superior feature extraction capabilities, outperforming existing methods in terms of performance and robustness.