基于NDSST的非平稳信号时频分析算法

应用时频分析方法研究含噪非平稳信号的时间-频率联合分布特性，De-shape SST（De-shape synchrosqueezing transform）算法具有良好的时频表现，但其抗噪性能及算法的鲁棒性还有待提高。提出基于非线性匹配追踪（nonlinear matching pursuit，NMP）分解的De-shape SST算法（nonlinear matching pursuit De-shape synchrosqueezing transform，NDSST），利用NMP良好的重构特性对非平稳信号进行稀疏重构，再进行De-shape SST时频分析，提高算法的抑制噪声能力和鲁棒性的同时，保留了良好的时频分布聚集度。数值仿真实验结果表明，对于单频、变频、线性调频和组合变频信号，NDSST算法可以得到高锐化度的时频分布表示，并且在低信噪比（sinal noise ratio，SNR）条件下依然具有优越的抗噪声性能。在金属破裂样本信号分析应用中，NDSST算法能够清晰地得出金属发生破裂的时间-频率范围，为工程实践中设置监测传感器的阈值提供判断依据。

Time-frequency Analysis of Non-Stationary Signal Based on NDSST

The time-frequency analysis methods are used to study the time-frequency distribution characteristics of non-stationary signal with noise. De-shape SST (De-shape synchrosqueezing transform) algorithm has good time-frequency performance. However, the anti-noise performance and robustness of the algorithm still need to be improved. A De-shape SST algorithm based on nonlinear matching pursuit decomposition (NDSST) is proposed. The sparse reconstruction of non-stationary signals is carried out via the excellent reconstruction feature of NMP, and then the time-frequency analysis by De-shape SST is performed. The approach in this article can improve the capability of noise reduction and robustness of the algorithm. Meanwhile, an accurate time-frequency aggregation is retained. The numerical simulation results show that the NDSST algorithm can obtain the time-frequency representation with high concentration for single frequency, variable frequency, linear frequency modulation and combined frequency conversion signal. And it has excellent anti-noise property in the low signal noise ratio (SNR) conditions. In the analysis and applications of metal rupture signal, the NDSST algorithm can clearly get the time and instantaneous frequency range of metal rupture. It provides a threshold value for setting up monitoring transducer in engineering practice.