Laboratory for Recognition and Organization of Speech
The Laboratory for Recognition and Organization of Speech and Audio investigates novel applications of signal processing and machine learning to extract information from audio signals including speech, music and the general sound environment.  I will describe some recent and ongoing projects in LabROSA, including:  - Eigenrhythms - subspace analysis of drum patterns extracted from pop music http://www.ee.columbia.edu/~dpwe/pubs/tr04-eigenrhythm.pdf  - Temporal envelope modeling via linear prediction applied to spectra, and applications in speech recognition and analysis/synthesis http://www.ee.columbia.edu/~dpwe/pubs/sapa04-plp2.pdf  - Segmenting and classifying long-duration `personal audio' recordings http://www.ee.columbia.edu/~dpwe/pubs/carpe04-minimpact.pdf  - Analysis and segmentation of recorded meetings using acoustic and lexical cues http://www.ee.columbia.edu/~dpwe/pubs/nist04-turnid.pdf http://www.ee.columbia.edu/~dpwe/pubs/nist04-laughs.pdf