Real-time monitoring of patients in hospital is of great importance, as it serves as an alarm of emergence condition. However, all-day company of carers or monitor is costly, and a waste of resources. With the development of deep learning, it is worthy of consideration to use low-cost real-time target recognition method in machine learning instead. This paper proposes to monitor the state of the patients via facial expression recognition. In order to that, a two-stage approach, i.e. detection of the face of the patient and classification the facial expression, is proposed. The face detector relies on the Harr feature, and is pre-trained. Then the detected face are classified either as “normal” or “abnormal” via a convolutional neural network. The training and test data are collected in real scene by mobile phone. The experimental results show an accuracy of 83% is achieved in test set.

Date Published: 27 November 2019
PDF: 5 pages
Proc. SPIE 11321, 2019 International Conference on Image and Video Processing, and Artificial Intelligence, 113210R (27 November 2019); doi: 10.1117/12.2547836

Published in SPIE Proceedings Vol. 11321:
2019 International Conference on Image and Video Processing, and Artificial Intelligence
Ruidan Su, Editor(s)