Students head-pose estimation using partially-latent mixture

 pose variation in the correct way from faces shots is an essential step for building interactive vision-based applications. In this work, a head-pose estimation system that predicts a head-pose angle for students during studying is proposed. The bounding-box method for face detection is used and a high-dimensional vector-space of faces using HOG features is adopted. A mixture of linear regression technique is applied to determine how to map high-dimensional vector-space onto bounding-box shifts and both head-pose parameters. The performance of the proposed system is evaluated and checked using the absolute error between the predicted angle and the target angle, after that, the mean absolute error (MAE) is computed along with standard deviation (STD).