Improvement Study of EEMD Decomposition Efficiency Based on CUDA Architecture

Abstract EEMD can inhibit mode mixing, which may occur in EMD, EEMD is a technology of adding many groups of white noise to original signal to do assisted analysis on the basis of EMD, however, it will greatly reduce the decomposition efficiency of the signal. In order to eliminate the effects of mode mixing, and improve decomposition efficiency of EEMD, this paper proposes a kind of EEMD based on GPU+CPU, the core idea of algorithm is to implement EMD decomposition’s iterative process control procedure by CPU, and the parts involved computing tasks are implemented by GPU. Decompose simulated signals and EEG respectively by using EEMD before and after improvement, and compares decomposition performance, the results show that, improved algorithm can greatly improve decomposition efficiency of EEMD.