Adaptive redundant speech streaming with scalable speech coding based on speech quality estimation

This paper proposes an adaptive redundant speech transmission (ARST) technique to improve the perceived speech quality (PSQ) of a speech streaming system over an IP network. To this end, the proposed technique estimates the PSQ from the received speech data under the current network condition. After that, it decides a suitable ARST mode in terms of the amount and bit rate of the redundant speech data (RSD) that can be used in a speech decoder to reconstruct lost speech signals for higher packet loss rates. On the basis of this decision, the proposed technique generates bitstreams of the current speech data (CSD) and the RSD using a scalable speech encoder to maintain the same transmission bandwidth. The effectiveness of the proposed technique is then demonstrated using ITU-T Recommendations G.729.1 and P.563 as a scalable speech coder and a PSQ estimator, respectively. It is shown from the experimental results that a speech streaming system employing the proposed ARST technique significantly improves the PSQ under various packet loss conditions.