Combining parametric and waveform-matching coders for low bit-rate speech coding

Waveform-matching coders preserve the shape of the target waveform and the time synchrony between the original and the synthesized signal. In parametric coders the shape of the encoded waveform is often changed, and the time-synchrony between the input and synthesized speech is not preserved. These two issues, time synchrony and waveform shape, are major obstacles in representing different speech regions with parametric and waveform coders, as arbitrary switching between the two results in annoying artifacts in transition regions. We describe a hybrid parametric/waveform coder with MELP used for strongly voiced regions and CELP employed for weakly voiced and unvoiced speech segments. To limit switching artifacts between the coders, alignment phase is estimated and transmitted in MELP making the original and synthesized speech time-synchronous. Additionally, in zero-phase equalization, the phase component of the CELP target signal is removed making the target waveform more similar to the MELP-synthesized speech. These two techniques, alignment-phase encoding and zero-phase equalization, greatly reduce switching artifacts in transition regions between the parametric and waveform coders. Formal listening tests of the 4 kb/s hybrid coder show that it can achieve speech quality equivalent to 32 kb/s ADPCM.