Transmission of multimode and variable-rate encoded speech samples on packet switched radio networks handling wide band voice information

Packet switching allows one to exploit the intrinsic variable-rate behaviour of real-time information, such as voice and video. Indeed, packet switching permits one to handle in a more flexible way the information bits and to obtain, therefore, a significant increment in system capacity. A relative measure of the capacity increment we obtain over circuit switched systems is the multiplexing gain. This paper presents preliminary results of the multiplexing gain we can achieve when we transmit multi-mode and variable-rate speech on a slotted and framed radio channel. For our evaluation we use an embedded, variable bit-rate speech coding scheme, namely the Mobile Audio Visual Terminal (MAVT) coder, candidate to the new standard MPEG-4. The peculiar characteristics of this coding scheme can be exploited for obtaining a significant multiplexing gain. Moreover, at high loads, a smooth and graceful degradation of speech quality can be obtained by exploiting the embedded property of the encoder. In fact, being that the information bits are classified by priority, we observe a graceful quality reduction during network congestion, if the transmission algorithm is designed to discard first the bits with the minimum impact on the output quality.