基于粒子群算法的欠定盲源分离方法改进研究

欠定盲源分离技术是一个热门的研究领域,其广泛应用于信息理论、神经网络、统计信号处理、生物医学工程等领域。在大多数实际情况下,当接收到由多路源信号叠加而成的观测信号时,源信号的数量大于观测时长,采用通常的盲源分离技术难以恢复源信号。着重讨论基于"两步法"的欠定盲源分离问题;该分离技术分两个阶段,第一阶段采用基于粒子群算法的K-均值聚类改进算法求解混合矩阵,将蚁群算法信息素的概念应用其中;第二阶段采用最短路径法求解L1-范数模型获得源信号的估计。相比于现存的二阶段方法,该方法可达到更高的信号重构信噪比。

The Improved Underdetermined Blind Source Separation based on Particle Swarm Optimization

Blind source separation is a new domain of signal processing, which involved information theory, neural networks, statistical signal processing, biomedical engineering and other fields. In most practical cases, the signals we received are multiplexed, which are a combination of several forms, and the number of source signals is greater than observed signals. The conventional technology of BSS is difficult to recover the source signals. The so-called underdetermined BSS is the case that the observed signal is not larger than the source signal, only by the observed signal can we recover the source signals. In this paper we focus on the two-stage underdetermined BSS, which consists two stages. In the first stage, the mixing matrix obtains by the K-means clustering algorithm, which based on particle swarm algorithm, and the concept of pheromone of ant colony algorithm is applied. In the second stage, in order to solve the L1-norm models, we apply the shortest path algorithm to obtain the estimated source signals. Compared to the existing two-stage methods, the proposed approach can achieve higher signal-to-noise ratio reconstruction.