Title :
A robust 1200 bps voice coder based on vector quantization of line spectral frequencies
Author :
Lu, Shan ; Baraniecki, Marion
Author_Institution :
COMSAT Lab., Clarksburg, MD, USA
Abstract :
This paper describes a 1200 bps voice codec that is based on a 10th-order linear prediction analysis, split vector quantization of line spectral frequencies, and differential pitch quantization. Robust codebooks and error protection and concealment techniques are used to minimize the effect of channel errors. Listening tests show that the codec has a high degree of speech intelligibility and speaker recognizability with natural voice quality
Keywords :
linear predictive coding; speaker recognition; speech codecs; speech coding; speech intelligibility; vector quantisation; vocoders; 10th-order linear prediction analysis; 1200 bit/s; channel errors; differential pitch quantization; error concealment; error protection; line spectral frequencies; listening tests; natural voice quality; robust codebooks; speaker recognizability; speech intelligibility; split vector quantization; voice codec; voice coder; Bandwidth; Books; Frequency; Linear predictive coding; Protection; Robustness; Speech analysis; Speech codecs; Speech coding; Vector quantization;
Conference_Titel :
MILCOM 97 Proceedings
Print_ISBN :
0-7803-4249-6
DOI :
10.1109/MILCOM.1997.648781
