论文部分内容阅读
用户交互行为是认知流媒体系统的基础和关键技术,将交互行为中的相关性规律建模为隐马尔可夫模型(hidden Markov model,HMM),并由此提出基于隐马尔可夫模型的流媒体数据预取策略.该策略使用Baum-Welch法对HMM的系统参数进行最大似然估计;然后基于HMM模型,利用当前用户的交互行为信息以及过去用户观看状态的后验概率进行贝叶斯推理,对用户当前观看状态的后验概率进行更新;最后根据最大后验概率准则对用户交互过程中的观看状态进行最终判决.使用后验概率,该策略可进一步确定具有最大预取价值的数据块,并实施预取策略以降低视频交互过程中的访问延迟.仿真实验证实了所提策略的有效性.
User interaction is the basic and key technology of cognitive streaming media system. The correlation law in interaction is modeled as hidden Markov model (HMM), and then the HMM (Hidden Markov Model) This strategy uses the Baum-Welch method to estimate the maximum likelihood of the system parameters of the HMM. Then based on the HMM model, using the current user’s interactive behavior information and the posterior probability of the past user viewing state, Bayesian Reasoning to update the posterior probability of the current viewing state of the user, and finally make a final judgment on the viewing state in the user interaction according to the maximum posterior probability criterion, which can further determine the data with the largest prefetch value by using the posterior probability Block and implement the prefetch strategy to reduce the access delay in the process of video interaction.The simulation results verify the validity of the proposed strategy.