Joint quantisation strategies for low bit-rate sinusoidal coding

Although there are speech coding standards producing high-quality speech above 4 kbps, below that transparent quality has not been achieved yet. There is still room for improvement at lower bit rates, especially at 2.4 kbps and below, which is an area of interest for military and security applications. Strategies for achieving high-quality speech using sinusoidal coding at very low bit rates are discussed. Previous work in the literature on combining several frames in a metaframe and performing variable bit allocation within the metaframe is extended. Experiments have been carried out to find an optimum metaframe size compromise between delay and quantisation gains. Metaframe classification and quantisation according to the metaframe class are used for better efficiency. A method for voicing determination from the linear prediction coefficient (LPC) shape is also presented. The proposed techniques have been applied to the SB-LPC vocoder to produce speech at 1.2 and 0.8 kbps, and compared to the original SB-LPC vocoder at 2.4/1.2 kbps as well as an established standard (Mixed Excitation Linear Predictive - MELP - vocoder) at 2.4/1.2/0.6 kbps in a listening test. It has been found that the proposed techniques have been effective in reducing the bit rate while not compromising the speech quality