基于多层非负矩阵分解的工频干扰消除

Removal of Power Line Interference Based on Multi-layer Non-negative Matrix Factorization

为保证信号测量可靠性和精确度,有效抑制工频干扰信号,提出了一种消除工频干扰的新方法。此方法避免了工频干扰的参数估计问题,也不需要单独输入参考源信号。以非负矩阵分解(NMF)为理论依据,以相似系数和重构信噪比为评估标准,进行工频干扰的消除。利用盲源分离思想,采用改进的多层NMF算法,对模拟数据和实测数据进行处理,有效完成了工频干扰信号的消除。

Power line interference is a common interference source in signal （such as bioelectric signal） with low frequency and weak amplitude detecting and transmission. It is an important technical problem to suppress power line interference effectively to gnarantee the reliability and accuracy of signal measure activities. One new method without complex parameter estimate and independent reference source signal was proposed to remove power line interference in this paper. The theory of non-negative matrix factorization （NMF） and the criteria of similar coefficient and signal to interference （SIR） were used to remove power line interference. Meanwhile, by employing the principle of blind source separation （BSS） and the improved algorithm of Multi-layer NMF into synthesized and real-life data, the method obtained satisfied and acceptable performance which is hard to be achieved by conventional filtering methods.