Modified Transductive Support Vector Machine for Efficient Facial Expression Recognition in Video

V. Sai Kumar and Dr.S.A.K. Jilani

Facial expression is in the area of dynamic research in the recent years. Detection and extracting various emotions and verification of those emotions from the facial expression turn into very significant in human computer interaction. Recently various methods have been introducing for developing automated facial expression recognition system. Those methods have major issue with speed and recognition results. To overcome this issue, Local Binary Pattern--Six Intersection Points (LBP-SIP) with Modified Transductive Support Vector Machine (MTSVM) approach is introduced for facial expression recognition. This work is used to track the dynamic changes in human face movements quickly to carry the important response system. In the proposed work, three dimensional (3D) video is processed as input for recognition. Haar features are extracted and face detection is carryout. Then LBP-SIP features are also extracted from detected faces. From the extracted LBP-SIP features the facial expressions are classified by using MTSVM approach. In MTSVM, the parameters are optimized by using Harmony Search (HS) algorithm. Finally the experimental results shows that the proposed works have better results than other existing systems in terms of accuracy, precision, recall and f-measure.

Volume 11 | 08-Special Issue

Pages: 1002-1011