Improved Pitch Modeling for Low Bit-Rate Speech Coders
Costantinos. Papacostantinou, P. Kabal · 1998
The poor modeling of the pitch pulse waveforms during voiced speech contributes to the degradation of CELP coded speech. This problem becomes more acute with the reduction of the number of pulses and other constraints im-posed on the xed codebook part of the excitation. We have developed a Pitch Pulse Averaging (PPA) algorithm to en-hance the periodicity of such segments, where during steady state voicing the pitch pulse waveforms in the excitation sig-nal evolve slowly in time. The PPA algorithm extracts a number of such pitch pulse waveforms from the past excita-tion, aligns them, and then averages them to produce a new pitch pulse waveform with reduced quantization noise. We have simulated and tested our algorithm on a floating point C-simulation of the G.729 8 kbps CS-ACELP coder. The results we present verify that the algorithm has generally improved the periodicity of voiced segments by reducing the average of the weighted mean-squared error. 1