Software and Data
We provide a Software Repository service for use by researchers in the UK audio and music fields:
We are also aiming to actively support and develop software to assist researchers. Software to be supported will be proposed by the audio and music research community, and is likely to include:
- Audio signal processing: denoising, de-echo, audio source separation, reference audio codecs
- Digital audio effects: filtering; time-stretching; timbre modification;
- Musical audio analysis: pitch tracking, chord analysis, key analysis, beat & rhythm analysis, fingerprinting
- Audio visualisation: time-frequency spectrograms, auditory models, structure analysis,
- Symbolic music tools: editing, typesetting, string pattern matching.
Software will be made available in cross-platform and in real-time versions for e.g. Max/MSP as appropriate. Some software for research on large-scale datasets will be required, for e.g. music information retrieval research on large databases of musical audio. For this we will initially build on our OMRAS2 framework to provide access to algorithms and data via a web services model.
Finally, we aim to provide a dataset repository to hold standard music & audio datasets, including:
- Musical audio datasets (esp. creative-commons);
- Audio source separation test-sets;
- Reference labels of chords, beats, notes, etc.;
- Symbolic music datasets.
Where copyright or licensing reasons mean that these cannot be made available freely, instructions on how to obtain the datasets will be maintained, with the Service holding a “reference” dataset for posterity in case of future unavailability from the originators. For music and audio data, as well as some software, copyright and licensing is an important issue. We will work with key rights holders to encourage availability of software and data for research purposes, so as to benefit research quality of UK and international research in this area.
