基于独立成分分析的表面肌电信号工频去噪

表面肌电信号(SEMG)采集中,如何消除工频干扰对信号的后续应用意义重大.在探讨独立成分分析(ICA)原理的基础上,提出了一种用于表面肌电信号工频去噪的快速独立成分分析(FastICA)算法.该方法通过对观测信号去均值和白化处理后,用负熵作判据通过迭代得到解混矩阵,经解混运算得到源信号.针对混合信号ICA分离效果的差异,引入最大似然指标作为分离效果的评价量.实验结果表明,所提算法能有效分离SEMG信号中的工频噪声,运用最大似然评价指标将工频噪声降至最低.

Power Frequency Noise-Reduction of SEMG Based on Independent Component Analysis

Power frequency noise-reduction in surface electromyography(SEMG) acquisition is very important for the subsequent application of SEMG. After a study on the independent component analysis(ICA),a neg-entropy-based FastICA method used to remove power frequency signal from SEMG is proposed. Before extracting each of the components, one needs to do some preprocessing to whiten the observed signals and remove the mean of the signals. The implementation of FastICA is seeking for a de-mixing matrix through the iterative algorithm based on neg-entropy criterion to get independent source signal. According to the difference of ICA separated result, the maximum likelihood value is proposed to evaluate the separated result. The experimental results indicate that the proposed method can efficiently remove the power frequency noise in the SEMG signals, and reduce power frequency noise to the great extent using the maximum likelihood value.