Can Software Tune Like Musicians?
Making Sampled Instruments Sound More Human Through Adaptive Tuning
Modern sample libraries of acoustic instruments have become very astonishingly realistic in terms of timbre and expression, yet most of the time experienced listeners can still distinguish them from live performances almost instantly. One possible reason is that sampled instruments remain tied to a fixed pitch grid, while human performers continuously adapt their intonation to musical context. In this talk we will be examining how an adaptive tuning system can be developed for sampled acoustic instruments which can closely replicate human intonation practices. The core engineering challenge of this project would be to determine how the software should determine pitch – Should it optimise frequency ratios? Should it adjust pitches based on harmonic context and voice hierarchies? Or should it attempt to minimise perceptual roughness? I will describe what existing adaptive tuning models attempt to optimise, and based on analyses of human performance data, attempt to firstly identify which optimisation principles explain human intonation most closely, and then to propose a direction for future adaptive tuning systems.
Yasandu Thewnula Kalansooriya Kalansooriyage
Student
University of the West of England, Bristol
Yasandu Kalansooriya (Yas) is an Electrical and Electronic Engineering Student at UWE Bristol, currently enrolled in his final year of study. His final year research focuses on developing an adaptive tuning system for sampled acoustic instruments that attempts to reproduce the intonation practices of real performers, with a long term aim of narrowing the perceptual gap between live ensemble performances and digital mock-ups.
This involves using Digital Signal Processing to analyse recorded ensemble performances across many different musical and acoustic characteristics and creating a deterministic model of human intonation behaviour.
Apart from engineering, Yas is very passionate about music in general, having learned piano at eight years of age and is now an advanced player, while being a casual composer as well. His broader interests include applying EE Engineering principles to the design of audio hardware and signal processing systems.