Investigation of an automatic voice intelligibility concept

A technique is described for electronically determining the degradation in intelligibility that occurs when speech is corrupted by added noise. Primary attention is given to the extraction of the relative time positions of the speech waveform zero crossings, the computation of correlation coefficients from the zero crossings, and the comparison of the coefficients with articulation scores at various ratios. The results of the investigation indicate that the technique is useful in providing a measure of intelligibility comparable to that obtained by a listening team.