Basic, algorithm of pitch synchronous innovation CELP (PSI-CELP) speech coding

This paper describes a new speech coding method called PSI-CELP (Pitch Synchronous Innovation CELP). This method is based on CELP and adds pitch synchronous innovation, which means that even random codevectors are adaptively converted to have pitch periodicity for voiced frames. PSI-CELP is suitable for low bit-rate speech coding and it can considerably improve the quality of synthesized speech without increasing the computational complexity or the bit-rate. In addition to pitch synchronous innovation, four other methods are used for low bit-rate CELP : (1) a fixed codebook that is prepared as part of the adaptive codebook for nonperiodic frames, (2) a random codebook consisting of two conjugate sub-codebooks, (3) an MA-type perceptual weighting filter using unquantized LPC parameters, and, (4) codebook arches using delayed decision. PSI-CELP improves the SNR, segmental SNR, and equivalent Q value of the coding system by 1.6dB, 1.0dB, and 0.5dB respectively.