Clinical infant pain trial based on k-NN algorithm

This paper presents a vision-based infant-pain monitoring system that adopts an infant behavior analysis approach to detect infant injuries. In our study, the system first pre-processes the input sequence to filter out the noise and reduce the effects of lights and shadows. Then, the infant's faces are detected from the input frames and feature extraction was done with SVD and FFT. A k-NN classifier was employed to describe pain over time. It is found that the identification rate of reaches 83.12%.