Poster Abstract: Toothbrushing Recognition using Neural Networks

Daily toothbrushing is essential for maintaining oral health. Recently, a wrist-watch based system was designed to monitor the effectiveness of toothbrushing at home. Since toothbrushing involves complex motions of the hand and the toothbrush, advanced machine learning techniques are needed for accurate recognition. In this paper, we design a neural network classifier to recognize toothbrushing surfaces based on wrist watch sensor data, including accelerometer, gyroscope and magnetometer data. Experiments shows that our algorithm successfully recognized toothbrushing surfaces with a recognition precision of 91.2%, outperforming existing solutions.