【作者】 陈永；

【导师】 赵力；

【作者基本信息】 东南大学 ， 信号与信息处理， 2006， 硕士

【摘要】 随着VoIP技术的不断发展和WLAN的日益成熟,VoIP技术和WLAN技术的融合是通信技术发展的必然趋势,可以预计WLAN上VoIP技术将是下一代通信领域中的关键技术之一。与传统电话相比,VoIP的致命的弱点就是语音质量较差。影响VoIP语音质量的因素是多方面的,关键因素之一是回声的影响。因此,要提高VoIP的语音质量,就必须在VoIP通信系统中进行消除回声的处理。目前通信系统中大多采用自适应回声抵消技术来实现回声消除。通信双方同时讲话(双端对讲)情况下的回声抵消是自适应回声抵消技术的主要难题之一。传统的自适应回声抵消技术都是通过在双端对讲情况下冻结滤波器系数调整的方法来解决这一问题。具体地说,就是通过信号检测技术确定通信双方的讲话状态,一旦检测到双端对讲,就禁止或放慢自适应滤波器系数的调整更新。这种基于双端对讲检测的自适应回声抵消技术会导致两个问题:一是由于自适应滤波器系数在双端对讲期间不调整,一旦回声路径在双端对讲期间发生变化,自适应滤波器就不能跟踪回声路径的变化,导致回声抵消效果下降;二是高性能的双端对讲检测算法大都比较复杂,双端对讲检测运算的延时性,有可能使自适应滤波器的系数被错误地冻结在已偏离最优解的滤波器系数上。回波抵消中的双端对讲检测器是用于监测远端和近端同时有话音存在的情况。它的作用是在近端有语音信号存在的时候停止滤波器的系数调整,以防止滤波器系数发散。许多文献资料中介绍了多种关于双端对讲检测的算法,其中包括能量检测法、互相关检测法和相干检测法,在这篇论文中我们将阐述这几种算法并辅以实验数据。更多还原

【Abstract】 With the fast development of VoIP (Voice over IP) and WLAN (Wireless Local Area Network) technology, the integration of WLAN with VoIP has become the mainstream for the future of communication. It is anticipated that VoIP over WLAN will be one of the most critical technologies employed in next generation communication system. Compared with traditional telephone, the fatal disadvantage of VoIP is its worse speech quality. The influence factors are various and one of the most important is the echo. So, in order to enhance the speech quality of VoIP, echo cancellation plays a critical role in the VoIP systems.At present, echo cancellation is mostly implemented by adaptive echo cancellation techniques. However, double talk, in which the near-end and the far-end speakers talk simultaneously, causes a serious problem in the adaptive estimation of the adaptive echo cancellation filter coefficients. To tackle this problem, almost all current adaptive echo cancellation techniques first detect the presence of double talk and then take appropriate actions such as disabling the adaptive algorithm during the double-talk period. But these techniques result in two problems: 1) Since the coefficients of the filter are not updated during the double-talk, once echo path changes during this period, it is impossible for the adaptive filter to trace the change of echo path and reduces the echo cancellation effect; 2) Most of these detectors are too complex in order to achieve satisfactory performance. Therefore, it is possible to make the coefficients of the filter diverge from the optimum due to the time required by the detector to detect double-talk.A doubletalk detector (DTD) is used with an echo canceler to sense when far-end speech is corrupted by near-end speech. Its role is to freeze the adaptation of the model filter when near-end speech is present in order to avoid divergence of the adaptive algorithm. Several authors have proposed to describe many different ways of doing so ,including the Giegel Algorithm, the cross-correlation algorithm and the coherence algorithm. We show in this paper that all the algorithms and the calculated data of the algorithm.更多还原