Patient State Recognition System for Healthcare Using Speech and Facial Expressions

J Med Syst. 2016 Dec;40(12):272. doi: 10.1007/s10916-016-0627-x. Epub 2016 Oct 18.

Abstract

Smart, interactive healthcare is necessary in the modern age. Several issues, such as accurate diagnosis, low-cost modeling, low-complexity design, seamless transmission, and sufficient storage, should be addressed while developing a complete healthcare framework. In this paper, we propose a patient state recognition system for the healthcare framework. We design the system in such a way that it provides good recognition accuracy, provides low-cost modeling, and is scalable. The system takes two main types of input, video and audio, which are captured in a multi-sensory environment. Speech and video input are processed separately during feature extraction and modeling; these two input modalities are merged at score level, where the scores are obtained from the models of different patients' states. For the experiments, 100 people were recruited to mimic a patient's states of normal, pain, and tensed. The experimental results show that the proposed system can achieve an average 98.2 % recognition accuracy.

Keywords: Local ternary pattern; Multi-directional regression; Patient state recognition; Smart healthcare.

MeSH terms

  • Facial Expression*
  • Humans
  • Information Systems / organization & administration
  • Needs Assessment*
  • Patients*
  • Pattern Recognition, Automated / methods*
  • Speech*
  • Time Factors
  • Video Recording / methods*