Investigation of Techniques for Evaluating Toothbrushing Performance Using Audio Data

This work explores the use of audio recognition techniques to evaluate how well users brush their teeth using only audio data collected from smartphones. We investigate two main approaches: a two-tiered approach that first performs activity recognition on the audio data and then uses the output of activity recognition to build a regression model for skill assessment, and an approach that uses deep neural networks to build a regression model for skill assessment that takes raw audio data as input.