SEGMENTATION AND CLASSIFICATION

Need to classify data into

• Speech
• Music
• Speech with Music

Segmentation Strategies

• BBN --- dual gender phone decoding, chop at pauses and gender turns
• CU-HTK --- frequency-based classification, gender-dependent phone decoding
• CU-Con --- syllable boundary information
• CMU --- GMM classifier, entropy-based segmentation at silences
• Dragon --- amplitude-based silence detection, phoneme recognition
• IBM --- GMM classifier, turn detection using BIC
• LIMSI --- GMM classifier, bottom-up agglomerative clustering
• OGI --- used CMU segmentations
• Philips --- frequency-based classification, gender-dependent phone decoding
• SRI --- context-dependent PTM classifier, bottom-up agglomerative clustering