A two level classifier process for audio segmentation

We are dealing in this paper with audio segmentation. We propose a two level segmentation process that enables the audio tracks to be sampled in short sequences which are classified into several classes. The segmentation is performed by computing several features for each audio sequence. These features are computed either on a complete audio segment or on a frame (set of samples) which is a subset of the audio segment. The proposed approach for microsegmentation of audio data consists of a combination of a K-means classifier at the segment level and of a multidimensional hidden Markov model system using the frame decomposition of the signal. A first classification is obtained using the K-means classifier and segment-based features. Then final result comes from the use of multidimensional hidden Markov models and frame-based features involving temporary results. Multidimensional hidden Markov models are an extension of classical hidden Markov model dedicated to multicomponents data. They are particularly adapted in our case where each audio segment can be characterized by several features of different nature.