FACIAL EXPRESSION RECOGNITION SYSTEM AND METHOD USING MACHINE LEARNING

摘要

The present invention relates to a facial expression recognition system using machine learning, and more particularly, to a facial expression recognition system, comprising: a detection module detecting a face region of a user from a video including a face of an input user; From the extracted face region, from the extraction module for extracting the feature vector for the user's face, the classification module for classifying the expression of the user using the feature vector extracted from the extraction module, and the user's facial expressions classified in the classification module, Including the recognition module for recognizing the facial expression of the user is characterized in its configuration. In addition, the present invention relates to a facial expression recognition method using machine learning, and more particularly, as a facial expression recognition method, (1) the detection module detects the face region of the user from the video including the input of the user's face (2) the extraction module extracting a feature vector for the face of the user from the face region detected by the detection module of the step (1), (3) the classification module, Classifying the facial expression of the user by using the feature vector extracted by the extraction module, and (4) recognizing, by the recognition module, the facial expression of the user from the facial expressions of the user classified by the classification module of step (3) It characterized by including the configuration. According to the facial expression recognition system and method using machine learning proposed by the present invention, a facial expression recognition system, which is detected by a detection module and a detection module for detecting a face region of a user from a video including an input of a user's face From the face area, the extraction module for extracting the feature vector for the user's face, the classification module for classifying the user's expression using the feature vector extracted in the extraction module, and the user's face from the facial expressions classified in the classification module And a recognition module for recognizing an expression, and the extraction module includes a landmark extractor and a feature vector extractor, and extracts a feature vector from angles and distance ratio information for each landmark from a landmark extracted from a user's face region. By reducing the amount of computation through minimizing the dimension of the feature vector, It can improve the speed of recognizing the user's facial expressions compared with face recognition technology. Further, according to the present invention, the classification module uses a random forest classifier and classifies the expression of the user into happiness, surprise, anger, neutral, and five other cases from the feature vectors extracted by the extraction module. The classification unit and the second classification unit which is classified into three cases of anger, disgust, and sadness when classified as other in the first classification unit, so that the random forest classifier trained using the learning data is included. Through the classification process of the step, it is possible to recognize the facial expression of the user in more detail.