Real-time Yawning Detection Based on Machine Learning Algorithm and Time Series Classification using Facial Feature Points

Yawning detection is one of the facial action recognition problems, which has a wide range of applications. However, traditional approaches in relevant fields have many shortcomings. This paper first discusses how to utilize OpenPose to capture facial feature points, and then puts forward a new yawning detection method LR+BOSS by integrating machine learning algorithm logistics regression and time series classification algorithm BOSS, based on the YawDD dataset. In the experiments, we demonstrate that such a strategy makes obvious improvements compared with the traditional method and has huge potential to be applied in the real-time yawn detection system.