Voice transmission over 802.11 wireless networks using analysis-by-synthesis packet classification

A new form of telephony is being made possible by the nearly-ubiquitous presence of 802.11 wireless local networks (WLAN's). This paper presents a technique to improve speech quality for WLAN telephony. Speech compressed by the 3GPP GSM AMR coding standard at 12.2 kb/s is packetized and classified according to an analysis-by-synthesis estimate of the perceptual importance of each individual packet. Perceptually important packets are protected against noise and channel collisions by a simple form of forward error correction, i.e., packet repetition. Network simulations and perceptual quality measures have been used to evaluate the performance of the proposed technique. Preliminary results show that protecting the perceptually most important 10% of all speech packets provides the same performance delivered by randomly protecting twice as many packets.