METHOD OF DATA REDUCTION OF SAMPLED SPEECH SIGNALS.

″Skipping″ of samples in uniform sample sequences can occur accidently or be purpously introduced. The recovery of skipped samples is of considerable importance in telecommunication. Resemblance of the recovered signal to the original bandlimited, continuous randomly generated signal has been shown to be strongly dependent on the nature of the skipping pattern. Much closer resemblance could be achieved with a derived random skipping pattern than with uniform skipping. This occurred at average sampling rates which were less than Nyquist rate. This sampling rate reduction is a form of data reduction. In this paper, the levels of the mean square errors of the recovered signal for uniform and nonuniform skipping is considered, together with the nature of the deviation of the reconstructed signal from the original computer simulated signal. The practical application of the results to signals, such as speech signals, is discussed.