DPCM with Forced Updating and Partial Correction of Transmission Errors
Robert JC Steele, D.J. Goodman, C. A. McGonegal · Bell System Technical Journal · 1979
A speech signal is dpcm (differential pulse code modulation) encoded at the sampling rate fsand also pcm encoded at a rate fs/W. Blocks of W dpcm words and one pcm word are transmitted. The receiver compares the decoded dpcm signal at the end of every block with the decoded pcm sample. If the difference is above a threshold, it is assumed that one error exists within the block, and a search is made for the erroneous dpcm code word. Correction is accomplished by inverting the bits in this code word until the difference is below the threshold. Whether or not the error is corrected, the dpcm signal at the end of the block is forced to the value of the pcm sample, thereby preventing error propagation outside the block. The system improves segmental s/n ratio by 7 dB for W = 64 and bit error rates between 0.1 and 0.5 percent Larger improvements are available with smaller block sizes. In the absence of transmission errors, there is no perceptible distortion due to the correction system.