摘要： 为解决传统算法训练的字典规模受限且运算量大的缺点,提出一种基于字典学习的语音信号欠定盲源分离算法,通过双重稀疏字典训练方法训练可稀疏表示的冗余字典并对观测信号进行稀疏分解。分析欠定盲源分离和压缩感知(CS)问题的等价性,构建基于CS的欠定盲源分离模型,并应用正交匹配追踪算法对信号进行重构,实现语音信号欠定盲源分离。实验结果表明,与KSVD算法和在线字典学习算法相比,该算法在保证分离精度几乎不变的前提下,能降低字典构建的计算复杂度,提高信号稀疏表示的有效性,并减少重构算法的运行时间。

关键词: 欠定盲源分离, 压缩感知, 稀疏表示, 冗余字典, 正交匹配追踪算法

Abstract: In order to overcome the shortcoming that the traditional learning algorithm training has limited dictionary size and large amount of computation,the algorithm of underdetermined blind source separation for speech signal based on the dictionary learning is studied.Firstly,a redundant dictionary is trained by adopting double sparse dictionary training algorithm to carry out the sparse decomposition of the observed signal.Then it analyzes the equivalence property of underdetermined blind source separation and Compressed Sensing(CS) equivalence problems,builds a model of underdetermined blind source separation based on CS,and also applies the Orthogonal Matching Pursuit(OMP) algorithm to reconstruct the signal to achieve speech signal underdetermined blind source separation.In the premise of guaranteeing the separation accuracy,the algorithm reduces the computational complexity of dictionary construction,improves the validity of signal sparse representation and reduces the running time of reconstruction algorithm.Experimental results show that the algorithm is better than the KSVD algorithm and online dictionary learning algorithm,which greatly improves the computational efficiency.

Key words: underdetermined blind source separation, Compressed Sensing(CS), sparse representation, redundant dictionary, Orthogonal Matching Pursuit(OMP) algorithm

中图分类号: 

• TN912.35