Hidden Markov modeling for network communication channels
Proceedings of the 2001 ACM SIGMETRICS international conference on Measurement and modeling of computer systems
Predicting packet loss statistics with hidden Markov models for FEC control
Computer Networks: The International Journal of Computer and Telecommunications Networking
Performance evaluation with hidden markov models
PERFORM'10 Proceedings of the 2010 IFIP WG 6.3/7.3 international conference on Performance Evaluation of Computer and Communication Systems: milestones and future challenges
Hi-index | 0.00 |
A number of applications can benefit from estimating future loss statistics. For instance, if the end-to-end loss characteristics of a path can be well approximated in advance, then a media streaming application could adapt its transmission parameters in order to deliver data with an acceptable quality to the user. In this work, we present a framework for adaptive prediction using hidden Markov models (HMMs). We propose a new class of hidden Markov models whose parameter values can be efficiently computed as compared to general HMMs. We also develop methods for predicting two measures of interest from HMMs, and perform experiments over a set of packet traces to assess the goodness of these predictions. Finally, we apply our prediction framework to dynamically select a forward error correction (FEC) scheme for media streaming. Using real Internet packet traces we evaluate the performance of our approach by emulating a VoIP tool. The PESQ algorithm is applied to assess the perceptual speech quality before and after the dynamic FEC selection. Our results show that the prediction-based approach achieves significant quality improvements with a small increase in the average transmission rate.