Text this: A markov-switching model approach to heart sound segmentation and classification