Noise reduction of sEMG by SVD based on neural network

According to the characteristic that the surface electromyogram signal (sEMG) is very weak and influenced by noise vulnerability, this paper proposes a new method that uses the unsupervised Kohonen neural network weights optimized to determine the order of the reconstruction matrix during the process of the noise reduction in singular value decomposition (SVD) effectively. First, let the sEMG collected through the Butterworth band-stop filter to remove 50Hz power line interference. Then use the SVD to deal with the signal filtered. To make use of the characteristic of the noise platform is gently and centralized of the singular value spectrum of the signal with noise. Through the projection on longitudinal axis in spectrum, we apply the Kohonen network optimized to confirm the boundaries of the noise platform, and then to determine the effective order of the reconstruction matrix. Simulation results show that this method achieve the noise reduction of sEMG preferably.