Mixed Type Audio Classification using Sinusoidal Parameters

A preprocessing stage in every audio application including music/speech separation, speech or speaker recognition and audio transcription task is inevitable to determine each frame belongs to which classes, namely: speech only, music only and finally mixture. Such classification can significantly lower the computational complexity due to factorial search commonly used in many model-based systems including monaural separation systems as well as music transcription scenarios. In this paper, employing sinusoidal parameters obtained by a fixed dimension modified sinusoidal model (FDMSM) already proposed in [7] a new classification approach is proposed to separate mixed type audio frames based on Support Vector Machine (SVM) and Relevance Vector Machine (RVM). The challenging problem in this work is seeking the most appropriate features to discriminate the underlying classes. As a result, we employ some unsupervised feature selection procedure to determine which feature to select to get the best results. The experimental results show that the proposed system presents acceptable classification result and outperforms other classification systems including k Nearest Neighbor (k-NN), Multi-Layer Perceptron (MLP).