Due to the lack of QoS mechanism in the present IP networks, the loss rate of real-audio packets will increase when network becomes more and more congested, and this will impact the audio quality to a great extent. Based on the study on the relationship between loss behavior of packets and the human perception of audio quality, an endto-end flow control algorithm is developed to reduce the side effects on the audio quality caused by lost packets. With the introduction of longterm and shortterm packets loss rate, the factors of human perception and network capacity are both taken into account in the algorithm, thus it carries out adaptation of audio coding rate more reasonably and exhibits better performance than what the other algorithms with the same purpose could achieve.
