Recognition of emotions in audio using transfer learning

Recognition of emotions in music is problematic task as it depends upon the state of mind of human. Music emotion detection is hot research topic which helps to know the psychological state of particular human being. However, various models have been deployed by using machine learning approach. In traditional machine learning approach we need to manually label large training dataset i.e. time consuming. To avoid implementing a module from scratch, transfer learning is used. Transfer learning is method were we can apply knowledge (gained while implementing domain related module) to target module. Audio features are extracted with limited set of training data and further classified by using convolution neural networks (CNN). In this paper we are proposing system which detects emotion in Indian classical music. Thus, by tagging limited set of data we can achieve higher accuracy result by using transfer learning approach.