Robust application of discrete all-pole modeling to sinusoidal transform coding

Current research in low bit-rate speech coding is showing increasing interest in sinusoidal modeling as used by sinusoidal transform coding (STC). Traditional STC requires an all-pole representation of spectra derived periodically from the short-term speech spectral envelope by peak-picking and cubic spline interpolation. In this paper, discrete all-pole (DAP) modeling is applied to STC to improve the accuracy of the short-term spectral envelope for voiced speech. The approach is made robust to occasional failure, for spectra which do not conform well to the all-pole model, by applying bandwidth widening to eliminate high frequency tonal artifacts. Objective distortion measures and informal listening tests have demonstrated that the robust application of DAP to STC can provide perceivable improvements in speech at bit-rates of 4 kbit/s and below