Multicomponent Signal for Comparing Direct and Indirect Methods of Speech Transmission Index Measurement

Abstract

When evaluating the intelligibility of speech distorted by noise and reverberation, direct or indirect methods of measuring the speech transmission index are used. However, it remains insufficiently studied how significantly differ the results of measurements obtained by direct and indirect methods. To find an answer to this question, the use of a multicomponent test signal consisting of four "elementary" signals separated by pauses is proposed in this paper. As "elementary" signals, it is proposed to use a maximum-length sequence, a speech shaped maximum-length sequence, a speech shaped stationary noise, and a speech shaped amplitude-modulated noise. Use of amplitude-modulated noise allows estimating speech transmission index by a direct method. Other "elementary" signals make it possible to estimate speech transmission index by two variants of indirect method. The proposed algorithms and corresponding computer programs were tested on trial signal models, while the consistency of the obtained results with the results of previous studies was revealed. The results of the signal models studies show that both considered variants of the indirect speech transmission index measurement method lead to underestimated results compared to the direct method. For one of the variants of the indirect method, the value of the estimate bias is 0.03–0.04, regardless of the interfering conditions. For another variant of the indirect method, the estimate bias varies from 0.01 to 0.18, depending on the interference conditions.