【作者】 张微；

【导师】 毛敏；

【作者基本信息】 华东师范大学 ， 通信与信息系统， 2007， 硕士

【摘要】 随着通信技术,计算机技术以及网络技术的不断发展,会议电话网建设也在高速进行。目前,会议电话技术正在向多媒体会议通信方向发展,所使用的网络也是目前发展最快的Internet,即可视化IP电话会议系统。而电话会议系统是利用计算机,多媒体通信技术和设备,通过IP网络传输信道在两地或者多个地点之间开会的一种通信手段。在一个电话会议系统中,音频模块起着非常关键的作用,衡量一个系统质量好坏的标准也往往在于此。在音频模块中,传统上是使用控制发言权的方法,即某一时刻只允许一个人发言。每个会议中首先选定一个会议主席,发言者在发言前必须向会议主席申请发言权,在发言结束后释放发言权。这样一来,大大限制了会议成员之间的交流,而采用混音技术即可解决电话会议中这样的问题。让更多的成员可以在同一时刻发言,并将所有发言者的信息传达给所有参与会议的成员,即所谓的混音。本文首先简单分析了VOIP的工作原理,几种基本的应用模式以及其关键技术。接着介绍了传统的以及一种自适应加权VOIP电话会议系统的混音算法,并在此基础上提出了一种改进后的适合于本系统的混音算法,此算法不仅在自适应加权算法的基础上进一步的提高了混音处理的速度,而且还引入了模拟电路中“柔性剪峰”(soft clipping)的思想,它能够对语音产生溢出的部分柔和切峰。经证明,这种算法的复杂度比自适应加权算法更低,在Windows XP系统,CPU为Celeron(R)/2.4GHz,内存256M的平台上,处理5路5秒的混音,本系统算法可比自适应加权算法省时0.4毫秒,当VOIP电话会议系统服务器同时为多个电话会议服务时,这0.4毫秒的优势就能够显而易见。另外,经改进后算法处理的混音质量比其他几种算法更高,这主要体现在语音的能懂度和清晰度上,采用将混音结果回放给没有参加录音的同学听,以调查问卷的形式反馈分析。由于终端的静音检测部分对混音的影响较大,所以课题还就静音检测方式进行了研究。鉴于本课题所研究的VOIP电话会议系统可支持固定电话、移动电话、软件电话等诸多种类的终端,所以最终依据传统的静音检测方法选择了一种改进了的,适合于本系统的短时能量算法以及过零率算法相结合的静音检测算法,只有当信号的短时能量以及过零率均小于阈值的情况下才判定为静音信号。更多还原

【Abstract】 With the development of the communication technology, computer technology and network, the creation of conference telephone network is developing fast. At present, the conference telephone technology moves forward to multimedia conference communication, which based on the internet, i.e. the visualized IP telephone conference system. The telephone conference system is a communication method between two or more locations via IP network, which uses computer, multimedia technology and apparatus. In the telephone conference system, the audio module is important, and is also a standard to measure a quality of a system.The audio module controls the locus say-so traditionally, i.e. only one person is allowed to speak in a moment. A chairman will be selected in a meeting first, and others have to applied to the chairman for the right to speak, and release the right after finishing his speech. Therefore, the communication of the members is limited tremendously, but the audio mix technology solves that problem. More members are able to speak in the same time, and deliver one’s information to all members, this is audio mixing.The passage simply introduces the operation principle of VOIP, some basic application mode and key technology. Then introduces an audio mixing algorithm of VOIP telephone conference system, and further bring up an advanced audio mixing algorithm that suits the said system. The algorithm is not only enhanced the audio mixing speed, but also apply the "soft clipping" in analog circuit, which can soft the overflow peak. It is proved that the said algorithm is easier than the self adapting weight algorithm. In windows XP system, CPU is Celeron(R)/2.4GHz, memory 256M, dealing with 5.5S audio mixing, the said algorithm saves 0.4 ms than the self adapting weight algorithm, when the VOIP telephone meeting system server serves for multiple telephone conference, the 0.4 ms advantage is quite obvious. In addition, the quality of the audio mixing is better than other algorithms, especially in understandable and clarity of the voice. The mix result can be replay to the members that have not attended the record, and feedback through the inquiry papers.As the voice active detect portion of the terminal affects the audio mixing obviously, so the voice active detect portion is also researched in this theme. The VOIP telephone conference system supports various terminals such as phones, mobile phones, software phones, so finally select an advanced mute check algorithm, which fits the short time energy algorithm as well as zero crossing algorithm. The signal is judged as mute only the short time signal and zero crossing rate are both below the threshold.更多还原