Droopy Mouth Detection Model in stroke warning

This paper presents a Droopy Mouth Detection Model in stroke warning. The objective of this paper is to take up the challenge to provide early detection of stroke through mouth drooping detection in mobile Android platform. To achieve that, a specialized library, Google Mobile Vision is utilized to detect facial landmark such as mouth corners and obtain the coordinates of the landmarks or key points for further processing. The inputs for the proposed droopy mouth detection model were taken from the Google Web Images and National Cheng Kung University (NCKU) Robotics Face datasets. The system prototype was evaluated using metrics such as Precision, Recall and F-Score to determine its recognition rate. Experimental results show that the proposed droopy mouth detection model has achieved satisfactory recognition rate. © 2016 IEEE.