“Human ‘Cross-Correlator’”—A Technique for Measuring Certain Parameters of Speech Perception

Abstract

A human listener is here regarded as a cross-correlator: his two ears are treated as input terminals, stimulated with the acoustic input signals f1 (t) and f2 (t+r), while his vocal responses are treated as the output correlation function. The two signals are, respectively, pure and distorted versions of the same signal. Delay is randomly set and the listener answers right or left, as the source of sound appears to him to lie. The correlation function then corresponds to the probability distribution of his correct judgments. Such functions represent the degree of aural fusion, and show up strikingly the invariants of speech signals which are significant in aural perception.