Human to human social communication in real-life is possible through different modalities like facial expressions, speech and body poses. However, facial expressions plays important role while dealing with human emotions in real-life than the other modalities. It is because facial expression provides non-verbal data towards emotions. And also gives emotion of a person towards his goal. On the other hand, speech and body poses are mostly language and culture dependent respectively which creates problem while detection emotions of a person. Thus in order to deal with the above issues, this research work focused on facial expressions instead other modalities. To improve detection performance of the system, proposed Relative Sub-Image Based features is used. Support Vector Machine with radial basis kernel is used for classification. Total six basic emotions (angry, sad, happy, disgust, boredom and surprise) are tested. From experimental results, the proposed Relative Sub-Image Based features enhanced the classification rates than the conventional features. © 2005 - 2016 JATIT & LLS. All rights reserved.