DCT-M model for the excitation parameter in low bit rate vocoders
Tang Kun · Journal of Tsinghua University(Science and Technology) · 2007
The description precision of the excitation signal greatly influences the quality of reconstructed speech in low bitrate vocoders.In order to improve the reconstruction quality,DCT-M model is proposed to express the excitation spectral parameter,which transforms the variable length vector to fixed dimension through 2-dimensional DCT transformation,and then quantize the fixed length vector using multi-stage vector quantization.Tests show that the method maintains the shape of the entire spectral envelope,reduces model errors and thus greatly improves the description precision.In the sine excitation linear prediction(SELP) vocoder,the DCT-M model improves the naturalness of the reconstructed speech,having subjective tests of 65%.