Local mean decomposition algorithm improved by de-correlation

Abstract

To solve the mode mixing problem of local mean decomposition (LMD), hereby a de-correlation improved LMD algorithm was proposed. If the multi-components signal includes two signal components with similar frequency, LMD will produce mode mixing which has serious impact on signal feature extraction and subsequent time frequency analysis. The essence of the mode mixing is that the information of product functions (PF) obtained by LMD mutual coupling each other. That is the PF is incomplete orthogonality. For the zero mean value random signal, the orthogonality and non-correlation are equivalent. By embedding the de-correlation operation in the LMD process, the orthogonality between the PF can be further guaranteed, and the purpose of suppressing the mode mixing is achieved. The simulation results show that the LMD algorithm improved by de-correlation has superior performance in suppressing the mode mixing.