Enhancing chord recognition accuracy using web resources
Matt McVicar, Tijl De Bie · 2010
Machine learning methods for chord recognition have improved considerably in the past few years. However, further progress seems constrained by the scarcity of training data. In this paper, we show that this problem can be partially solved by exploiting noisy but freely and abundantly available online resources, in addition to fully labeled training data. We use these data to restrict the output of the Viterbi algorithm, resulting in significant improvements over the standard decoding process.