Towards complex multi-component pulse signal with strong noise: Deconvolution and time-frequency assisted mode decomposition

Pulse signal is a kind of signal with large amplitude in a short time, which widely exists in the real world, such as vibration signals of rotating machinery, EEG signals and ECG signals, etc. The analysis of pulse signal is of great significance to reveal the change laws of corresponding objects. Affected by various factors, the frequency and amplitude of these pulse signals are complex and are disturbed by noise. It is challenging to analyze pulse signals with strong noise. To solve this problem, we propose a novel deconvolution and time-frequency assisted mode decomposition (DTMD). DTMD mainly consists of two core parts, namely signal deconvolution and signal decomposition. DTMD first establishes a new fast deconvolution method to filter the noise. This deconvolution method optimizes the sparse constraint function, and adopts the idea of signal local update optimization, which can filter the noise of pulse signal quickly and effectively. Then, integrating signal demodulation optimization with signal's time-frequency feature information, DTMD build a new time-frequency supported signal decomposition model, which can effectively highlight and protect the instantaneous frequency ridge information to accurately decompose pulse signals. In this way, DTMD can accurately decompose complex pulse signals with strong noise, and effectively reveal the pulse signal's change property. The processing experiments of several complex simulated and experimental pulse signals show that the proposed DTMD can decompose complex pulse signals with strong noise more precisely than the existing signal decomposition and TFT methods. The proposed DTMD can effectively filter the strong noise, and accurately estimate the IFs of pulse signal. Therefore, the proposed DTMD has strong practicability.