Mac7: Adaptive Smart Toothbrush

Dental caries has a prevalence of more than 90%. To prevent cavities, smart toothbrushes can help to guide the user and give valuable feedback. This requires the concise localization of the toothbrush within the mouth of the user. As of today, there is no convenient technique available to detect the tooth and surface brushed concisely. Therefore, this paper presents a prototype of a smart toothbrush including a mobile application and a novel approach to solve this issue based on the brushing style of the user. We utilized a 3-axis accelerometer, 3-axis gyroscope, 3-axis magnetometer combined with the random forest machine learning algorithm. With an accuracy of 99.63% for the right-handed model and an accuracy of 99.71% for the left-handed model, an evaluation of the localization detection approach showed an average accuracy of 9.0%. This shows that the brushing habits might be too variable to be captured by the random forest algorithm or that the data are not sufficient to generalize the pattern.