Music emotion recognition (MER) plays an important role in human to machine interaction and digital entertainment. Emotions conveyed by any given music track is a challenging task that has received considerable attention in the music information retrieval. A various efforts have been made towards the music emotion recognition. Among from the regular features used in MER, there is a search for new features for recognising the music emotions. Residual phase feature is an excitation source feature is used in this work can be used to exploit emotion specific information present in the music signal. In this work MER categorizes emotions into five classes: anger, fear, happy, neutral and sad. The features extraction and classification plays an important role in MER, in such a way they represents music well and computation can be carried out efficiently. MFCC is the well known timbre features combined with residual features are used with pattern classification techniques like autoassociative neural network, support vector machine and radial basis function neural network. The performance of the method was measured using equal error rate and recognition rate.