Abstract
A variable-length packetization process is proposed for logarithmic Pulse Code Modulation (log-PCM)-encoded speech signals. Blocks of W log-PCM words are reduced in size by discarding words on the basis that the receiver can recover the speech with a signal-to-noise ratio (s/n) that is, in general, above a specified value. Specifically, blocks of two hundred fifty-six, 8-bit, μ-law PCM words are left unchanged or reduced to either 214, 205, 192, 171, or 128 words. These blocks are then formulated into packets. The discarded samples at the dispatching terminal are replaced at the receiver by means of adaptive interpolation. We found that by specifying the s/n of the decoded speech in each variable-length packet to be above 27 dB, the reduction in the transmitted speech data was 25 percent, while the recovered speech signal had negligible perceptual degradation.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
