A Solution to the Bandwidth Compression Problem--Important Points
A. M. Sandman, J. Endresen · IRE Transactions on Communications Systems · 1976
The basic problem with all previous methods of bandwidth compression has been that the accuracy of the reconstructed waveform has been poorly controlled. By reconstructing the wavefrom from the compressed digital signal before transmission and creating the error waveform, i.e., substracting the reconstructed from the original waveform to produce the error waveform, followed by adding points to be transmitted, where the error waveform is greater than specified, a process which may be iterated, the residual error in the final reconstructed signal may be reduced to any arbitrary amount. Unlike "run-length coding" only the minimum number of points necessary to reconstruct the waveform are generated. By introducing a time delay, considerable bandwidth compression would be possible while keeping predetermined accuracy specifications.