A novel deep learning optimization algorithm for human motions anomaly detection

Recently, there are many researches to detect the anomaly of human motions by using the machine learning and inertial sensors. In general, the individual differences exist in human motion by the height, body weight, habits and so on. Classification models based on the deep learning have often high quality. However, the general deep learning optimization algorithms do not consider the individual differences in human motions. By the reason, classification model based on the algorithms does not guarantee to take into account the individual differences. Therefore, we propose a novel deep learning optimization algorithm for human motions’ anomaly detection from the data of the inertial sensor. The reliability of the proposed algorithm is also confirmed by the collected dataset.