Reducing the complexity and storage of CELP speech coding using a self-orthogonal codebook

An algorithm is proposed to reduce the complexity and memory requirement of coded-excited linear prediction (CELP) speech coding. The new algorithm is based on the concept of designing a special codebook such that each codeword is orthogonal to its shifting entries. With this orthogonal property, the algorithm reduces the codeword searching complexity of CELP coding significantly. Besides, by rearranging the codeword, only 12.5% of the conventional codebook storage is required. Both segmental SNR and informal listening showed that the performance of the algorithm is equivalent to that of the original CELP algorithm.