当IP语音包的网络时延抖动较小时，一般的语音缓冲算法可以得到较好的语音质量。当网络中存在突发大时延时，就会出现极大丢包率或极大端到端时延，从而难以获得好的语音质量。为此，提出针对突发大时延下的自适应语音缓冲算法。通过估算网络平均时延和学习语音包经过的网络路径上的状态，来确定需要控制端到端时延大小和语音包的丢包率，动态调整Jitter Buffer队列的最小深度和最大深度，从而可以尽量减小语音裂缝（gap）的出现。通过基于听觉模型的客观音质评价（PESQ）仿真计算以及在实际语音网关设备中的应用表明算法对语音通信质量有一定的改善作用。-The continuous playout of voice packets in the presence of variable network delays is often achieved by buffering the received voice packets for sufficient time. Basic jitter buffering algorithms can work well only when the delay does not spike status of the networks, is presented to promote the quality of voice communication. It timely adjusts the minimal and maximal depth of buffer queue according to the control target of end-to-end delay and packet loss rate. The algorithm can much more easily achieve the continuous playout because it plays voice packet at a fixed inter-play time in the most time of a talk-spurt. The control target of packet loss rate can be extended to 20 . However, the basic algorithms can only bear 5-10　of the packet loss rate. Perceptual evaluation of speech quality(PESQ) is applied to assess the speech quality in the simulation. It is shown that the algorithm can obviously promote the quality of voice communication in IP networks with spike delay. The practic