Key points are not available for this paper at this time.
This method makes use of the fact that the voiced portion of a speech signal is very nearly periodic in time, the period being the inverse of the fundamental frequency. If a periodic wave is delayed by exactly one period and then subtracted from the original wave, the result is zero. If the delay is other than a multiple of the period, the result is in general not zero. To implement this principle, speech is passed through a tapped delay line and the output of each tap is subtracted from the undelayed signal. If the voltages thus obtained are time averaged and then scanned, a wave results which shows a null at the tap corresponding to the period. In the case of unvoiced sounds a null will not appear, so that voiced and unvoiced sounds can be differentiated.
Miller et al. (Sun,) studied this question.