Does 16-QAM provide an alternative to a half-rate GSM speech codec?

The computational complexity, speech quality, spectral efficiency, and robustness tradeoffs of speech transmission schemes for personal communications networks (PCNs) are addressed. Differentially encoded star 16-QAM arrangements with and without oversampling and diversity are compared, in terms of robustness, to the parallel frequency-division multiplex (FDM) 16-quadrature amplitude modulation (QAM) system to find the most appropriate modem scheme. The 13.4-kb/s regular-pulse-excited-long-term predictor (RPE-LTP) speech codec is subjected to rigorous bit-sensitivity analysis in terms of both cepstral distance (CD) and segmental signal to noise ratio degradations. A twin-class embedded binary Bose-Chaudhuri-Hocqunghem bit protection scheme is devised to improve robustness. The 6.2-kBd system proposed guarantees low-complexity, high-quality speech transmissions in a bandwidth of 12 kHz for vehicular speeds of 30 mph and channel signal to noise ratios in excess of 25 dB, a value readily maintained in a microcellular PCN