Integrating networks measurements and speech quality subjective scores for control purposes

Traditionally, QoS has been addressed by using network measurements (e.g., loss rates and delays), and little attention has been paid to the quality perceived by end-users of the applications running over the network. Here, we address the issue of integrating speech quality subjective scores and network parameters measurements, for designing control algorithms that would yield the best QoS that could be delivered under a given communications network situation. First, we build a neural network based automaton to measure speech quality in real time, at the style of a group of human subjects when participating in an MOS test. We consider the effects of changes in network parameters (e.g., packetization interval, packet loss rate and their pattern distribution) and encoding on speech signals transmitted over the network. Our database includes transmitted speech signals in different languages. Then, we outline a control mechanism which, based on the application performance within a session (i.e., MOS speech quality scores generated by the neural networks), dynamically adjusts parameters (codec and packetization interval). Finally, we analyze preliminary results to show two main benefits: first, a better use of bandwidth, and second, delivery of the best possible speech quality given the network current situation.