表情認識のための顔部品の特定とその抽出

概要

論文の詳細を見る
This paper propose a new algorithm for extracting facial parts such as mouth. After extracting facial area, the facialparts are extracted for the recognition of facial expressions. In general, Q-element in a YIQ color system is used forextracting facial parts. The difficulty in extracting facial parts is how to identify the positions of these facial parts beforeextracting them. In the previous methods, a single threshold value was set for identifying each facial part. In many faces,however, more than one of facial parts are extracted as the candidate facial parts. After extracting the candidate facialparts, therefore, the actual positions of facial parts must be selected by examining their relative positions. In orderto identify each facial position, the most appropriate threshold value is set individually for each facial part in the proposalmethod. After the each position of facial part is identified, the actual image of facial parts are extracted based on templatematching. The template used here is created automatically by examining texture and edges of each facial part.
2012-10-30