Human emotion recognition based on remote video sequencing

点击翻译按钮获取中文摘要

Emotion modeling and recognition has drawn extensive attention from disciplines such as psychology, cognitive science, and, lately, engineering. The objective of this Engage grant is to develop and test at the University of Toronto an automated video based human emotion recognition system that can be deployed in a flexible and cost effective manner. The supporting company Architech Labs has developed such methodologies for affective analysis of video sequences that can reveal the user's emotional state. However, there is significant demand for ubiquitous, enterprise level deployment on a variety of devices. This presents a new challenge for emotion recognition systems since recognition has to be performed with uncontrolled lighting and pose conditions as well as other restrictions of a real world setting. The specific application that we will focus on is developing emotion recognition (ER) methodologies to classify audience engagement vs boredom for educational applications. The envisioned system would be able to provide a teacher (or a presenter) with real -time audience engagement feedback. There are all sorts of modalities that one can use for that but, in essence, the heart rate, pupil blinks and perhaps gaze could convey interesting information for this application. All data will be collected remotely by video sequencing using Architech's ER technologies and UofT's affective computing and video processing expertise.