I’m working on Cover Version Retrieval, an audio-based music information retrieval project that identifies different cover performances of the same composition despite changes in instrumentation, arrangement, tempo, and other musical characteristics. Its technical focus includes audio signal processing and feature extraction, Harmonic Pitch Class Profiles (HPCP), similarity measurement, Dynamic Time Warping (DTW), alignment paths, and evaluation.
I’ve been validating the repository’s implementation against its published results, extracting numerical data from experimental figures, and checking that its DTW implementation reproduces published alignment paths. The Python test suite was independently verified to contain 119 passing tests, compared with the 112 tests mentioned in the original README. I’m interested in extending this work toward more sophisticated music understanding and audio representations. The project is on GitHub at https://github.com/hey-shiv/cover-version-retrieval.
I independently verified that the Cover Version Retrieval Python test suite had 119 passing tests.
The original README mentioned 112 tests for Cover Version Retrieval.
I said I have been working with music, audio, and AI, and that I am also a classical musician.