Assessing the stability of the spectro-temporal cue weighting angle for listener categorization
Sara Ng, Gregory Matthew Ellis, Pamela E. Souza, Frederick J. Gallun, RICHARD A. WRIGHT, Mari Ostendorf · Proceedings of meetings on acoustics · 2021
The cue profile procedure developed by Souza et al. (2015) provides a reliable quantified measure of listeners’ use of spectral and temporal speech cues. This study provides an analysis of the time course stability of the cue weighting angle calculation and suggests a modification of the cue weighting angle calculation that accounts for potential instability of the time course. The results suggest that a significant reduction may be made to the number of trials needed to compute the cue weighting angle. Simple machine learning categorization provides improves cue weighting angle prediction using fewer trials.