| The industry makes a consensus to adopt packet-based IP as NGN technological basis and thus enable network operable and manageable to transmit future communication service. At present, there are lots of multimedia services over the Internet. But there is no perfect QoS over IP network, and high delay, delay jitter and packet-loss are the features of the Internet - the biggest IP network. These may badly degrade the speech quality, especially for VoIP - the realtime and interactive speech session. Some technologies such as DiffServ, MPLS have been applied to improve network performance. On the other hand, the terminals over the Internet have some intelligent i.e. powerful processing capabilities. This intelligent make it possible to solve the inherent problems and general deficiencies introduced by IP network.In this paper, the features of packets transmit over the Internet and real-time, interactive audio session are investigated firstly. Then based on the study on audio capturing, encoding, decoding, transmitting and speech enhancement a media stack used in softphone are designed and implemented.Several crucial parts in media stack of VoIP are studied and applied such as adaptive playout buffer, voice activity detection and combination of network technology and signal processing technology. The media stack can tolerate network jitter and packet loss. Some advanced features such as discontinuous transmit also are implemented to reduce the transmission rate during inactive speech periods while maintaining an acceptable level of output quality. Whenever over network simulation or over the Internet, the media stack can provide well audio communication. At present, it has been applied to SIPHello– a softphone based SIP. |
