一种新的基于稀疏表示的单通道盲源分离算法

发布时间：2018-07-02 23:09

本文选题：稀疏表示 + 单通道盲源分离　；参考：《电子与信息学报》2017年06期

【摘要】：该文针对稀疏表示应用于单通道盲源分离中存在字典间互干扰的问题,通过在常规联合字典中引入一个新的子字典——"共同子字典",提出一种新的基于稀疏表示的单通道盲源分离算法。新的字典学习目标函数中单个源的保真度由对应子字典和共同子字典构成,共同子字典的存在可以有效避免某一源信号在其他子字典上寻求成份而带来的互干扰问题。目标函数的求解通过交替执行稀疏表示、字典更新和比例系数优化3个步骤来实现。在测试阶段,通过收集单个源所对应子字典和共同子字典上的分量可以估计出混合信号中的单个源信号,从而达到盲源分离的目的。在语音数据库上进行的对比实验发现,所提算法较传统算法和前沿算法在两个通用评价指标上最高有近1 dB的提高。
[Abstract]:In this paper, the problem of dictionary interferences in single channel blind source separation based on sparse representation is discussed. A new single-channel blind source separation algorithm based on sparse representation is proposed by introducing a new sub-dictionary, "common sub-dictionary", into the conventional joint dictionary. The fidelity of a single source in a new dictionary learning objective function is composed of a corresponding sub-dictionary and a common sub-dictionary. The existence of a common sub-dictionary can effectively avoid the problem of mutual interference caused by a source signal seeking components in other sub-dictionaries. The solution of the objective function is realized by alternating sparse representation, dictionary updating and proportion coefficient optimization. In the testing stage, the single source signal in the mixed signal can be estimated by collecting the components of the corresponding sub-dictionary and common sub-dictionary of a single source, thus achieving the purpose of blind source separation. The comparison experiments on the speech database show that the proposed algorithm is better than the traditional algorithm and the forward algorithm in two general evaluation indexes, which are up to 1 dB.
【作者单位】：空军工程大学航空航天工程学院;
【基金】：国家自然科学基金(61372167) 航空科学基金(20152096019)~~
【分类号】：TN911.7

【相似文献】