SVD-Based Filtering of Noise in Harmonics with Wavelet Preprocessing

In this paper we performed assessment of both wavelet and SVD based techniques and proposed a hybrid method by combining wavelet and SVD for the filtering of noise in harmonics. In the first step of the algorithm, wavelet denoising is applied to the noisy signal as pre-processing (gross filtering) step. Singular value decomposition is then applied to a data matrix formed from the denoised signal. The filtered signal is recovered from the signal truncated SVD subspace of this data matrix. Performance of each scheme has been evaluated by simulation on the basis of relative mean square error.