Research On Vision-Based Emotion Cognition Of Home Service Robot

In the field of harmonious human-computer interaction and artifical intelligence, affective computing as a new research direction has attracted increasing attention. Using computer vision method, we let home service robot cognitive human emotion by recognizing facial expression, and make a corresponding response.Firstly, we use AdaBoost algorithm to detect human face and three kinds of organs on it, that is eyes, nose and mouth. This object detection method firstly presented by Paul Viola and Michael Jones is real-time, robust and achieves high detection rates. Detecting human face is used for collecting facial expression database and locating face image to be recognized, while detecting the other three kinds of organs is used for transforming the in-plane rotated face image to up-right. Compared with human face containing abundant features, eyes, nose, mouth and other facial organs contain incomplete features, we can't efficiently detect them in complex background, so we detect facial organs in the face area.After that, we use principle component analysis to extract facial expression features from face image, and construct facial expression database in feature space for following facial expression recognition. While locating facial expression area, we don't simply consider face area as facial expression area, in fact we locate facial expression area more precisely by using the geometric information of human face. Our innovation and main contribution are reconstructing the feature database inspired by distance-based hashing presented by Vassilis Athitsos et al. This scales the size of feature database, and raises efficiency of facial expression recognition.Finally, we use K-nearest neighbor algorithm to classify static image or image frame picked from camera for facial expression recognition. After recognizing facial expression, home service robot will take action accordingly to interact with human in reason.