论文部分内容阅读
本文提出了一种基于最近邻线分类器的新的双端检测器(DTD)。主要的思想是充分地利用特征信息以及用模式识别方法来设计DTD。本文从模式分类的角度分析了二种主要的传统DTD(Geigel和相关DTD)并给出了新的设计方法。一种称为NNL分类器的新的非参数分类器被用来检测双端通话。NNL分类器具有低运算量和优良的性能。用NNL分类器,我们熔合了几种传统的DTD并且避免了存在于大多数传统DTD中的固定阈值带来的问题。因此NNL-DTD在各种条件下是鲁棒的。实验结果也显示出了这个方法比传统方法更有效。
This paper presents a new double-ended detector (DTD) based on nearest neighbor classifier. The main idea is to make full use of feature information and pattern recognition methods to design DTD. This paper analyzes two main traditional DTDs (Geigel and related DTDs) from the perspective of pattern classification and presents new design methods. A new non-parametric classifier called an NNL classifier is used to detect double talk. NNL classifier with low computational complexity and excellent performance. With the NNL classifier, we fuse several traditional DTDs and avoid the problems posed by fixed thresholds that exist in most traditional DTDs. Therefore NNL-DTD is robust under a variety of conditions. The experimental results also show that this method is more effective than the traditional method.