Current lines of work
Symbolic music genre recognition
This is one challenging problem in music information retrieval and music style perception understanding. It has a number of applications, like the indexation and exploration of music databases. Statistical techniques are being applied to digital scores for solving this task.
Track identification in multi-track MIDI files
Automatically identifying the track containing the melody line in a MIDI file is a not well studied task with applications in music retrieval and melodic comparison. The approach can be extended to other instrumental parts (like bass or piano) or any other concept (lead or accompaniment parts).
Automatic transcription of digital audio
The automatic extraction of the notes in a digital musical signal permits both applying symbolic processing algorithms to audio data and score obtaining and processing. The polyphonic and polytimbral tasks are open problems in signal processing research. The role of human feedback in interactions and multimodality are being studied.
Automatic music analysis
Melodic, harmonic, and functional analysis are open problems by themselves but they are also powerful tools for helping in other tasks like music segmentation, key identification and tracking, music comparison, chordal progressions, reductions, performance rendering, etc.
Music similarity metrics
This is a well-know task in music information retrieval, since it permits music comparisons. The need of good metrics is a key point in this problem. Data structures suitable to represent music data and are explored in our group. Also probabilistic apporaches to similarity are being explored.
One of the classical tasks in computer music and artificial intelligence has always been composition. Our background in music representation and cognition is also used for style-guided composition through evolutionary methods and pattern recognition approaches.
Paper in Music-AI Workshop, IJCAI 2007. Demos.
Digital sound synthesis
Techniques for using computers to sound generation, involving computer languages like Csound. Control sequencies generation for performance rendering and music composition.
Music encoding and cultural heritage
Grammars and libraries for encoding and decoding different symbolic music languages. In particular, we are interested in coding ancient music with the capability of designing transducers able to translate old notations to modern music representations.
"Plain and easie code" ANTLR v4 grammar