欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
在信源个数未知的语音信号盲分离中，首先要解决的是信源个数的估计问题。在深入分析基于判定主特征值个数的信源数估计方法出错原因和条件的基础上，提出利用对少量观测样本自相关矩阵的特征值进行聚类分析的方法来实时估计语音信源数。同时，依据期望聚类过程，定义了一种新的类间距离，并给出了特征值聚类过程的详细计算步骤。仿真实验表明该方法的估计准确率明显优于基于主特征值的估计方法。
英文摘要：
The accurate estimation of number of sources plays a key role in the blind speech source separation. The dominant eigenvalue-based method often results in the wrong estimation of the number of speech sources due to nonstationary properties of the speech. Therefore, an eigenvalue clustering-based estimation method for the number of speech sources is proposed based on a new distance among different classes defined by the expected process of clustering. Simulations show that the method can more accurately estimate the number of speech sources than the dominant eigenvalue-based method.
同期刊论文项目
同项目期刊论文
Adaptively combined FIR and functional artificial link neural network equalizer for nonlinear commun
Functional link neural network cascaded with Chebyshev orthogonal polynomial for nonlinear channel e
期刊信息
位置：