Enhancing Public Speaking Skills Through AI-Powered Analysis And Feedback

Abstract

Public speaking anxiety, or glossophobia, affects a significant portion of the population, hindering effective communication. Existing solutions, such as traditional coaching and generic online courses, often fail to provide personalized, realtime feedback. These approaches lack the ability to dynamically analyze both verbal and non-verbal cues in a holistic manner. This paper presents an AI-driven application designed to enhance public speaking skills through personalized feedback on voice modulation, facial expressions, and speech content. By employing technologies such as SpaCy and NLTK for text processing, OpenCV and YOLO for facial expression and gesture recognition, and OpenAI-Whisper and SpeechRecognition for speech analysis, the system provides users with targeted, actionable insights to improve their performance. Experiments conducted using a custom dataset containing videos of speeches demonstrate an overall system accuracy of 87.73%, with individual component accuracies of 92.25% for text processing, 76.45% for facial and gesture recognition, and 92.5% for speech analysis.