In order to discover the relationship between music and the emotion that it may evoke, twenty-one features have been extracted to describe music. A feature selection algorithm called sequential floating forward selection (SFFS) is utilized to find discriminative features. An estimation of the correlation coefficient was applied to determine features of music that evoke an emotion. These features were then used to train two support vector machines (SVMs) for an individual subject to classify music that evokes happiness, anger, sadness, and peacefulness. Experimental results show that the proposed approach can be used to classify music that evokes an emotion for an individual subject with high classification accuracy.