Fast VQ-CELP-a speech coder for mobile radio
T.J. Moulsley, P.W. Elliott
Abstract
T.J. Moulsley, P.W. Elliott
Abstract
The currently emerging generation of mobile radio systems employs digital speech transmission techniques to achieve higher system capacity than analogue systems. Therefore methods of transmitting high quality digital speech signals at low bit rates are of great interest in such applications. The CELP (code excited linear prediction) algorithm is a good basis for meeting this requirement. However, in its original form CELP is too computationally complex for implementation in portable equipment with current technology. The authors present a novel modification of the CELP architecture (fast vector quantisation CELP) which greatly reduces the computational complexity while maintaining good speech quality. A codec based on the new algorithm has been submitted as a candidate for the 1/2 rate GSM standard, which is intended to double the capacity of the GSM system by halving the gross bit rate used for speech transmission. >
A significance statement is not available in the OpenAlex record.
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.
The currently emerging generation of mobile radio systems employs digital speech transmission techniques to achieve higher system capacity than analogue systems. Therefore methods of transmitting high quality digital speech signals at low bit rates are of great interest in such applications. The CELP (code excited linear prediction) algorithm is a good basis for meeting this requirement. However, in its original form CELP is too computationally complex for implementation in portable equipment with current technology. The authors present a novel modification of the CELP architecture (fast vector quantisation CELP) which greatly reduces the computational complexity while maintaining good speech quality. A codec based on the new algorithm has been submitted as a candidate for the 1/2 rate GSM standard, which is intended to double the capacity of the GSM system by halving the gross bit rate used for speech transmission. >
Key concepts: Code-excited linear prediction, Computer science, Speech coding, Vector sum excited linear prediction, Full Rate, GSM, Speech recognition, Codec2