Abstract: This paper analyzes main factors affecting VoIP(Voice over Internet Protocol) quality. It calculates network parameters such as delay and loss rate based on RTCP, and then maps them to different speech quality scores. According to the scores, voice terminals’ bit rate is adapted to network congestion status with guarantee of speech quality. Experimental results show that the method is effective for Internet congestion reduction

Key words: RTCP; Flow control; Speech quality evaluation; RBF neural networks

摘要： 分析了影响因特网语音通信质量的主要因素，根据实时传输控制协议计算延迟、丢包率等网络参数，并利用RBF 神经网络将参数映射为语音质量评分，根据评分值的变化动态地调整语音终端的编码速率，在保证一定语音质量的同时，缓解网络拥塞。实验证明了该方法的有效性。

关键词: 实时传输控制协议；流量控制；语音质量评价；RBF 神经网络