Machine Listening in Interactive Music Systems: Current State and Future Directions.
Artemi-Maria Gioti · University of Michigan Library Repository · 2017
This paper compares different machine listening strategies in interactive performance and improvisation systems. A three-stage model of interactive music systems is proposed, consisting of a sensory processing, a decisionmaking and an action module. In the sensory processing stage, feature extraction (analysis) functions as the basis for the construction of higher-level representations of the human input (interpretation).The types of data involved in this stage include MIDI-level, spectral and gestural or textural information, while analysis techniques can range from data processing (sorting, classification, data smoothing, statistics etc.) to real-time machine learning.A broad spectrum of interactive systems are discussed and compared with respect to their machine listening algorithms, their degree of autonomy and their application in composed or improvised music.