Smiling faces are better for face recognition

TitleSmiling faces are better for face recognition
Publication TypeConference Papers
Year of Publication2002
AuthorsYacoob Y, Davis LS
Conference NameFifth IEEE International Conference on Automatic Face and Gesture Recognition, 2002. Proceedings
Date Published2002/05//
ISBN Number0-7695-1602-5
Keywordsbetween-class scatter matrices, Databases, discrimination power measure, dynamic scenarios, expressive faces, face recognition, facial expressions, performance, performance differences, smiling faces, software performance evaluation, Training, visual databases, within-class scatter matrices

This paper investigates face recognition during facial expressions. While face expressions have been treated as an adverse factor in standard face recognition approaches, our research suggests that, if a system has a choice in the selection of faces to use in training and recognition, its best performance would be obtained on faces displaying expressions. Naturally, smiling faces are the most prevalent (among expressive faces) for both training and recognition in dynamic scenarios. We employ a measure of discrimination power that is computed from between-class and within-class scatter matrices. Two databases are used to show the performance differences on different sets of faces