Conversational quality as a function of delay and interactivity

The impact of mouth-to-ear delay on Conversational Quality (CQ) has been subject to manyfold research in the field of telecommunications. However, its dependency on Conversational Interactivity (CI) has not been studied explicitly in a quantitative way. To investigate the impact of both delay and CI, we conducted 48 tests with 96 different participants and controlled the interactivity through different tasks, namely natural conversation (flight), proof reading of text (text), and random number verification (rnv). The control of interactivity is proven to be effective and remains stable for delays from 100 ms to 800 ms. The Speaker Alternation Rate (SAR) is used as the metric for interactivity. It is shown to have a significant impact on CQ after 200ms delay and the effect does not alleviate until 800 ms. The hypothesis is proven using both graphical analysis and statistical evaluation based on a linear model and Analysis of Variance (ANOVA).