A digital coding and its application to speech transmission is described. The coder determines the amplitudes and times of sucessive extremes (relative maxima and minima) of the signal. This information is decoded at the receiver by interpolating a function between extremes so as to connect them smoothly and preserve the extremes of the original signal in the reconstructed wave, Thus, the coding is a nonlinear sampling technique. It is related to clipped speech encoding which effectively transmits only the times of the extremes. The properties of the coding for speech signals have been studied by digital simulation on an IBM 704 computer. Information rate, statistics of the extremes data, and quality of the resulting signal have been evaluated. The buffer size necessary to receive the randomly occurring data and transmit at a constant rate was measured.

A coding scheme is described for the transmission ofncontinuous correlated signals overmchannels,mbeing equal to or less thann. Each of themsignals is a linear combination of thenoriginal signals. The coefficients of this linear transformation, which constitute anm \times nmatrix, are constants of the coding scheme. For the purpose of decoding, themsignals are once more combined linearly intonoutput signals which approximate the input signals. The coefficients of the coding matrix which minimize the sum of the mean square differences between the original signals and the reconstructed ones are shown to be the components of the eigenvectors of the matrix of the correlation coefficients of the original signals. The decoding matrix is the transpose of the coding matrix. As an example, the coding scheme is applied to a channel vocoder in which speech is transmitted by means of a set of signals proportional to the speech energy in the various frequency bands. These signals are strongly correlated, and the coding results in a substantial reduction in the number of signals necessary to transmit highly articulate speech. The coding theory can be extended to include the minimization of the expectation of any positive definite quadratic function of the differences between the original and reconstructed signals. In addition, if the signals are Gaussian, the sum of the channel capacities necessary to transmit the transformed signals is shown to be equal to or less than that necessary to transmit the original signals.