论文部分内容阅读
利用lp模约束的稀疏成分分析方法可以对信号进行去噪处理。基于lp模约束的稀疏分解方法通常是采用优化方法来对信号在过完备库中进行分解,而分解时采用的过完备库是非常巨大的,所以如果优化方法选择不当,会导致稀疏分解效率的低下。在本文中,采用BFGS方法来进行优化分解,和通常采用的Newton方法比较,能在保持稀疏分解结果性能基本不变的前提下,有效地提高算法的分解效率。
The signal is denoised using sparse component analysis using lp-mode constraints. The sparse decomposition method based on lp-modulo constraint usually adopts the optimization method to decompose the signal in the overcomplete library, and the overcomplete library used in the decomposition is very huge, so if the optimization method is not chosen properly, the sparse decomposition efficiency low. In this paper, the BFGS method is used to optimize the decomposition. Compared with the commonly used Newton method, the decomposition efficiency of the algorithm can be effectively improved while keeping the performance of the sparse decomposition result substantially unchanged.