DocumentCode :
2873991
Title :
8Kbit/s LD-aCELP Speech Coding with Backward Pitch Detection
Author :
Shuhong, Wu ; Gang, Zhang
Volume :
2
fYear :
2009
fDate :
18-19 July 2009
Firstpage :
434
Lastpage :
437
Abstract :
This paper presents an 8 Kbit/s speech coding algorithm whose delay is 2.5 ms on the base of G.728 algorithm to reduce coding rate. Adaptive codebook structure is introduced in the proposed algorithm, which is composed of the recent excitation. The algorithm uses double codebook structure which contains adaptive codebook and normalized fixed codebook. In codebook searching, the algorithm detects backward pitch, and then searches the adaptive codebook subtly round the pitch period T. After searching 64 adaptive code words, 8 gains, 256 fixed code words and 8 gains, the best excitation is obtained. Testing with the average segment SNR and perceptual evaluation of speech quality (PESQ), the improved algorithm is close to G.728 in speech coding quality. And it achieves desirable balance in coding rate, delay time and coding quality.
Keywords :
acoustic signal detection; adaptive codes; linear predictive coding; speech coding; speech intelligibility; G.728 algorithm; LD-CELP speech coding algorithm; PESQ algorithm; adaptive codebook structure searching; average segment SNR; backward pitch detection; delay time; double codebook structure; low-delay code-excited linear prediction; normalized fixed codebook; perceptual-evaluation-of-speech-quality; Adaptive filters; Delay effects; Educational institutions; Information processing; Paper technology; Quality of service; Software algorithms; Speech analysis; Speech coding; Speech synthesis; Adaptive codebook; Backward pitch detection; Low Delay Code Excited Linear Prediction (LD-CELP); Perceptual Evaluation of Speech Quality (PESQ);
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Information Processing, 2009. APCIP 2009. Asia-Pacific Conference on
Conference_Location :
Shenzhen
Print_ISBN :
978-0-7695-3699-6
Type :
conf
DOI :
10.1109/APCIP.2009.243
Filename :
5197230
Link To Document :
بازگشت