ANALYSIS AND QUANTIZATION OF GLOTTAL PULSE SHAPES

In source-filter based speech coding for low bit rates an efficient re presentation of excitation pulses is required to attain high quality o f the synthetic speech. In this paper, we discuss a pulse waveform rep resentation by a codebook populated with pulse shapes. The codebook is designed from glottal derivative pulses obtained by a linear predicti ve inverse filtering technique. Pulses are extracted and normalized in time and amplitude to form prototype pulses. Design methods and perfo rmance evaluation of the codebooks are investigated in a vector quanti zation (VQ) framework. The quantization gains obtained by exploiting t he correlation between pulses are studied by theoretic calculations wh ich suggest that about 2 bits per vector (in a budget of 7-10 bits) ca n be gained when exploiting the correlation. Memory based VQ is a gene ric term for quantization schemes which utilizes previous quantized pu lses. We study traditional memory based VQ methods and an extension of memory based VQ with memoryless VQ, denoted a safety-net extension. T he experiments show that performance improves when extending memory ba sed VQ with a safety-net. It is found that, at the designated bit rate s, a safety-net extended memory based VQ can gain about 1.5-2 bits in comparison with memoryless VQ. (C) 1998 Elsevier Science B.V. All righ ts reserved.