A NEW FACIAL EXPRESSION RECOGNITION TECHNIQUE USING 2-D DCT AND K-MEANS ALGORITHM

ogiiitioii plays )I vital rolt: ill rmliiit,elligi:111 IIIIIII~I~I-III~ICI~~I~~: iiit.erlace, a,iid ttlacted IIIIICII attont;ion. 111 t,liis paper, we propsi~ a iiew facial osprvssion rw,gnit,iou trret,lio(l tllat iitilizes ttie 2-D DCT, k-nieaits ;ilgorit,tnn and vector oratcliing. Tliis t,ecliniqiir is b;u;ed on two iliain intiiitive ideas: (i) ~:un~plicnted facial exprmioti categories SIICII as “anger” and “sadness”, may bc: divided into sevcral subcnl.egorii:s wit,lr dillireirt~ suh featurc spaces wkiere the recognition task call he performed with higher accuracy, and (ii) the k-irrea,ns algorithni may hc nsed t,o cluster these snbcategories. A new image database with five facial expressions (neutral, smile, anger, sadness, surprise) of 60 women was constructed using a conrputat-ionally efficient projection-based teclrnique. Fxperimental resiilts using the new database and an existing one (60 men) reveal that the new technique outperforms the standard vector matching technique and two recently developed methods using fixed-size and constructive one-hidden-layer neural networks. The mean recognition rate can be as high as 95% for the two databases.