A variable length packetization scheme is proposed for packetizing blocks of W A-law PCM speech data at network nodes.When voiced speech occurs a variable length packetization scheme discards every n-th sample,n=2,3,4,5or ∞,such that when these rejected samples are replaced at the receiver by means of adaptive interpolation the interpolation noise is below a specified value,The value of n determines the length of speech data in the packet.Only the magnitude of the spectral components,Alaw PCM encoded,are transmitted for unvoiced speech.Packets representing the silence condition contain no data.The accumulative reduction in data was close to 50% For two concatenated sentences when the signal-tointerpolation noise ratio per block exceeded 25dB,and W=256.
展开▼
