High quality multi-rate CELP speech coding for wireless ATM networks

The paper presents a multi-rate CELP speech coder which meets the requirements of high perceptive quality, robustness to noisy environments and flexibility required by speech communications in the wireless ATM scenario. The codec, exploiting new robust algorithms for multilevel phonetic classification and efficient activity/inactivity speech coding models, presents an average bit rate of 4 kbit/s with an algorithmic delay of 15 ms and a perceptive quality better than that of the GSM full-rate 13 kbit/s codec and similar to that of the 8 kbit/s ITU-T 0.729 standard, both in the discontinuous transmission mode. In addition, the paper demonstrates that the multi-rate speech coder proposed is a promising coding approach for future wireless ATM-based networks in that it exploits the bit rate variability within talkspurts, thus guaranteeing, with the same average bit rate as an on-off speech coder, greater robustness to packet loss and therefore a better quality of service (QoS).