摘要
在分析基于BP神经网络的多路复合协议数据单元(MUX-PDU)提取算法及其局限性的基础上,根据H.223协议中mobile level 2模式下的MUX-PDU后验信息,提出一种基于BP神经网络及后验信息的联合MUX-PDU提取算法。该算法可动态调整MUX-PDU标识符的判决门限,同时在发生标识符虚警或漏报的情况下,能利用后验信息对提取出的MUX-PDU帧进行校验修正。仿真结果表明,该算法可有效降低MUX-PDU的丢包率,为提高3G视频通话质量提供了可靠保障。
On the basis of an analysis in the MUX-PDU extracting algorithm based on only BP neural network and its limitations, according to the posterior information of MUX-PDU under mobile level-2 in H.223 protocol, a united MUX-PDU extracting algorithm based on BP neural network and posterior information is proposed. This algorithm is able to adjust the decision threshold of MUX-PDU identifier dynamically, and making corrections to the extracted MUX-PDU frames by using posterior information is capable. Simulation result shows that this algorithm is able to decrease the packet loss rate efficiently, which provides a reliable guarantee for video calls in the 3G communication.
出处
《计算机工程》
CAS
CSCD
北大核心
2011年第1期276-279,共4页
Computer Engineering
