文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Design and implementation of a multicast audio conferencing tool for a collaborative comput
侵权投诉

Design and implementation of a multicast audio conferencing tool for a collaborative comput

We present the design of CCFAudio, a collaborative audio tool, and report some experiences with its usage. The tool is specifically designed for collaboration, supporting very high audio quality, non-Mbone multicast, mixing, and hands-free operation. 1

of these use the Mbone for multicasting packets whereas we use our own UDP IP-multicast. This avoids congesting the Mbone and enables users without Mbone access to collaborate1. Free Phone is the more complete tool of the two, including such features as a phonebook and the ability to\call" someone. The audio tool in 4] attempts to improve audio quality by re-sending the same audio sample in two consecutive packets but in di erent encodings. This repair scheme has been generalized in 5] where an audio packet can contain several audio samples in di erent speech codings. CCFAudio currently has no such repair mechanism, but does have hooks to implement the scheme in 4]. However, we have found by experiment that even with 20% packet loss, conversation is quite understandable via CCFAudio.

Storage buffer

send-audio

Audio source saudio

Multicast

Figure 1: Send-Audio Subsystem

this method naively results in poor audio quality. To conserve bandwidth, CCFAudio transmits only when the user is speaking, while avoiding dropouts when the user is too quiet to trigger voice detection. A number of ad hoc measures have been implemented to this end. A sample with volume less than v is considered 3 The CCFAudio Tool silent; otherwise it is treated as voice. In both CCFAudio is an audio conferencing tool with adcases the sample is bu ered. ditional features tailored for use in a collaborative session. Considerable e ort was spent to achieve If the current sample contains voice and the audio quality comparable to telephone (conference) previous sample did not, then up to k1 bu ered calls. CCFAudio tool has evolved into the means of samples that have not been transmitted are communication among members of the CCF team. sent prior to sending the current sample. A common use of CCFAudio is debugging other CSCW applications under development by describUp to k2 samples are sent after the last voice ing to one another what each is seeing on the local sample. display. CCFAudio consists of two audio subsystems and We have found empirically that the latter two a graphical user interface (GUI) subsystem, each measures can improve audio quality signi cantly. running as a separate thread or threads. The rst However, the parameters v, k1 and k2 are highly subsystem,\send-audio", pa

cketizes voice samples sensitive to the articulation of individual speaker from a source and transmits them to other tools. and the audio hardware used. CCFAudio therefore The second subsystem,\receive-audio", receives allows a user to change these values dynamically packetized voice samples from multiple sources, using its graphical interface. Feedback from the mixes the samples and plays the result. The GUI recipients is most helpful in determining the best subsystem provides a point and click interface for values. controlling the audio parameters as well as graphi- Send-audio can also accept input from an audio cal feedback showing who is in the conference and le. In this case, the saudio thread will read and transmit data at the data rate given in the header who is speaking at any time. information of the audio le. Figure 1 shows the send-audio subsystem. There is only one thread, saudio, in send-audio, that repeatedly reads a xed amount of data from the audio source and stores it into a bu er. If the sample contains voice data then it is converted to the user-speci ed format (e.g., GSM) and multicast to all members in the session. Implementing

3.1 The Send-Audio Subsystem

3.2 The Receive-Audio Subsystem

Human hearing is very sensitive to interruption of the audio stream, either as gaps in the stream, or as dropouts. Frequent interruption of the stream can make a conversation unintelligible. It is therefore very important to ensure that the audio output device does not become overloaded due to over ow 1 We discovered, to our surprise, that many natural sci- or idle due to under ow. (Over ow occurs when entists do not have access to the Mbone. data is written to the audio output device faster 2

上一页第2页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email