Primitive audio genre classification: An investigation of feature vector optimization

In this paper, we propose a content-based audio classification approach to classify audio signals into primitive genres such as speech, music, speech with music, and speech with noise. We analyze different characteristic features of audio signals in time, frequency, and coefficient domains and select me optimal feature vector by employing an analytical method to each feature. In order to ensure automatic classification by a fuzzy c-means (FCM) algorithm, we utilize a hybrid classification framework by combining FCM with k-nearest neighbor (KNN) algorithm and apply it on the extracted normalized optimal feature vector to achieve an efficient classification result. Experimental results on a robust dataset demonstrate that the proposed approach outperforms the existing state-of-the-art audio classification approaches by more than 10% of accuracy improvement in audio genre classification.