Overview

This repository contains the rule-based probabilistic algorithmic music generation system presented in:

Ehrlich, S. K., Agres, K. R., Guan, C., & Cheng, G. (2019).
A closed-loop, music-based brain-computer interface for emotion mediation.
PLOS ONE, 14(3), e0213516. https://doi.org/10.1371/journal.pone.0213516

The system generates continuous, controllable affective music in real time. Its key property is that it allows smooth transitions between affective expressions (e.g., happy → sad → angry) by continuously varying two control parameters:

• val ∈ [0, 1] (valence)
• aro ∈ [0, 1] (arousal)

These affect parameters are mapped to musically meaningful structural parameters, enabling music generation along arbitrary trajectories through valence–arousal affective space (Russell/Thayer style).

---

Algorithm in brief

Core concept

The algorithm is implemented as a rule-based probabilistic state machine that generates streams of MIDI events (notes/chords). The probability of note events and their attributes is continuously modulated by the affect controls (val, aro).

Mapping emotion → musical structure

Following established psychological findings on music emotion, the system maps valence/arousal onto five structural music parameters.

1. Tempo (linked to arousal)
2. Rhythmic roughness / density (linked to arousal; more notes = higher complexity/activation)
3. Relative loudness of subsequent tones (linked to arousal)
4. Pitch register (overall pitch) (linked to valence)
5. Harmonic mode (linked to valence; discrete 7-step diatonic modes (Lydian → … → Locrian) ordering)

Musical style / rendering

The generator outputs MIDI and was used in the study with three virtual instruments:

• piano
• cello
• bass

resulting in a small classical chamber-orchestra style. :contentReference[oaicite:6]{index=6}

---

Repository structure

code-algorithmicMusicGenerationSystem/
├── composer_algorithm.m            # main probabilistic MIDI generator + affect mapping (val/aro → structure)
├── trajectory_gen.m                # generate trajectories through affective space (valence/arousal over time)
├── documentation.pdf               # mathematical formulation of the composing algorithm
├── trj_*.mat                       # example affective trajectories (valence/arousal)
├── sound_examples/                 # rendered audio examples of generated music
├── russel.png                      # affective space illustration (valence/arousal plane)
├── LICENSE
└── README.md

Citation

If you use this code in research, please cite:

@article{ehrlich2019closedloop, title = {A closed-loop, music-based brain-computer interface for emotion mediation}, author = {Ehrlich, Stefan K. and Agres, Kat R. and Guan, Cuntai and Cheng, Gernot}, journal = {PLOS ONE}, volume = {14}, number = {3}, pages = {e0213516}, year = {2019}, doi = {10.1371/journal.pone.0213516} }