A framework for classification and segmentation of massive audio data streams