Audio Fingerprinting​

Machine Learning for Signal Processing - Carnegie Mellon University (2009)

The project was mainly to provide a web service that people can record and upload a short recording of certain music to find out the name of the song, such as Shazam. We have calculated STFT to create constellation map from local peaks of log power spectrum and then generated fingerprints based on local maximas. Since the fingerprints were saved through combinatorial hashing, we were able to match fingerprints quickly.

Algorithm Overview

Algorithm Details

Publications

Lam, A.; Chae, H.; Walsh, R.; Thompson, S. (2009). Audio Fingerprinting. Paper presented internally at Carnegie Mellon University.