This is a hands-on lab/seminar, which complements the technologies and methodologies covered by the core courses of the Master's program. The course focuses on a multimodal approach to audio and music information analysis, from varied sources (eg. audio, score, text, web) and dataset sizes (from single snippets to massive collections). It complements the mainstream, 'engineering-only', music information retrieval approaches, by introducing methodologies proper from humanities (musicology, creativity, critical argumentation). The course, thus, explores the cross-fertilizing relations between musicology, music perception, and music computing, from a user-centered standpoint.
- Musicological, perceptual and computational foundations of musical analysis
- 'Big-data' audio and music information mining
- Multimodal audio and music information processing (audio, video, score, text, user tags, motion capture, biosignals, ...)
- Musical analysis in the audio domain
- Musical analysis in the symbolic domain: music notation, MIDI, ...
- Musical analysis in real time: evaluation-by-sonification
- Musical analysis-by-composing: generation and transformation
- Multimodal-related signal processing: audio-to-score alignment, audio source separation
Class format and evaluation
Lab sessions consist of short lectures introducing few concepts/problems, followed by hands-on activity, which may span the rest of the week as homework. Students are expected to present their work/progress to the classmates, as well as participate in the discussions.
Evaluation is based on the following items:
- Lab reports (70%)
- Class presentations (20%)
- Active participation in the discussions (10%)