Dual Rate Low Delay CELP Coding (8kbits/s 16kbits/s) using a Mixed Backward / Forward Adaptive LPC Prediction
Stéphane Proust, Claude Lamblin, Dominique Massaloux
Abstract
Stéphane Proust, Claude Lamblin, Dominique Massaloux
Abstract
This paper describes a dual rate 8kb/s / I6kb/s low delq CELP speech coder with a 5ms frame length. Our studies have focused on the design of a classified structure based on both a long term (voiced / unvoiced) and a short term segmentation of the speech signal. This classifcation enables to combine a backward and forward LPC structure in a same codec. The bit allocation can then be better optimized according to the characteristics of rhe input signal and the desired bit rate. Results of formal tests are presented and show that the quality is equivalent to the G729 ITV-T codec at 8kb/s (preliminary version of October 94) and outperforms significantly G728 at 16kb/s.
OpenAlex reports 4 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
This paper describes a dual rate 8kb/s / I6kb/s low delq CELP speech coder with a 5ms frame length. Our studies have focused on the design of a classified structure based on both a long term (voiced / unvoiced) and a short term segmentation of the speech signal. This classifcation enables to combine a backward and forward LPC structure in a same codec. The bit allocation can then be better optimized according to the characteristics of rhe input signal and the desired bit rate. Results of formal tests are presented and show that the quality is equivalent to the G729 ITV-T codec at 8kb/s (preliminary version of October 94) and outperforms significantly G728 at 16kb/s.
Key concepts: Code-excited linear prediction, Computer science, Speech coding, Full Rate, Codec, Adaptive Multi-Rate audio codec, Codec2, Linear predictive coding