Real-time Public Speaking Anxiety Prediction Model for Oral Presentations

Abstract

Oral presentation skills are essential for most people's academic and career development. However, due to public speaking anxiety, many people find oral presentations challenging and often avoid them to the detriment of their careers. Public speaking anxiety interventions that help presenters manage their anxiety as it occurs during a presentation can help many presenters. In this paper, we present a model for assessing public speaking anxiety during a presentation - a first step towards developing real-time anxiety interventions. We present our method for ground truth data collection and the results of neural network models for real-time anxiety detection using audio data. Our results show that using an LSTM model we can predict moments of speaking anxiety during a presentation.