摘要
An adaptive denoising algorithm based on local sparse representation （local SR） is proposed. The basic idea is applying SR locally to clusters of signals embedded in a high-dimensional space of delayed coordinates. The clusters of signals are represented by the sparse linear combinations of atoms depending on the nature of the signal. The algorithm is applied to noisy chaotic signals denoising for testing its performance. In comparison with recently reported leading alternative denoising algorithms such as kernel principle component analysis （Kernel PCA）, local independent component analysis （local ICA）, local PCA, and wavelet shrinkage （WS）, the proposed algorithm is more efficient.
An adaptive denoising algorithm based on local sparse representation （local SR） is proposed. The basic idea is applying SR locally to clusters of signals embedded in a high-dimensional space of delayed coordinates. The clusters of signals are represented by the sparse linear combinations of atoms depending on the nature of the signal. The algorithm is applied to noisy chaotic signals denoising for testing its performance. In comparison with recently reported leading alternative denoising algorithms such as kernel principle component analysis （Kernel PCA）, local independent component analysis （local ICA）, local PCA, and wavelet shrinkage （WS）, the proposed algorithm is more efficient.
基金
Supported by the National Natural Science Foundation of China under Grant Nos 60572025 and 60872123, the Joint Fund of the National Natural Science Foundation and the Guangdong Provincial Natural Science Foundation,China,under Grant U0835001.The authors thank S. Q. Zheng and her research team for sharing their software for the purpose of comparison.
