Abstract: Singular Value Decomposition (SVD) is a widely used nonlinear signal denoising approach. Its denoising effect is depending on the selection of singular values for the signal reconstitution. The singular value difference spectrum of singular value can be used to select the expected singular values adaptively. However, sometimes when dealing with signals with strong trend, such as aircraft engine health signals, it may hardly choose the right singular values and fail to restore the desired signals To solve such problem, a signal denoising approach based on Empirical Mode Decomposition (EMD) and SVD is proced. First, the original signal is decomposed by EMD and the trend component is extracted from the original signal. Then, SVD is used to denoise the detrended signal, and the difference spectrum of singular value is used to select the singular values adaptively. After denoising, the detrended part is added to the trend part to get the final signal. Results of Both simulation and real aircraft engine health signal experiment demonstrate that the proposed approach can precisely select the singular values for signal reconstitution and effectively reduce the noise.

Key words: information processing, signal denoising, singular value decomposition(SVD), empirical mode decomposition(EMD), aircraft engine