An Efficient Time Domain Speech Compression Technique and Hardware Implementation on TMS320C5416 Digital Signal Processor

摘要

Speech compression is of paramount importance in modern day communications where bandwidth conservation is necessary to accomodate the ever increasing number of communication channels. Historically, LPC (Linear Predictive Coding) is used for speech analysis for better results, and subband coding is used in audio compression methods, such as MPEG standards. In this paper, we combine a RELP (Residual Excited Linear Prediction) technique with subband coding to yield compression rates of upto 85% with acceptable distortion levels.