Tuesday, September 24, 2019, 9:00
Building 106, 4th Floor, Room 04 007
In the music industry, sample libraries are digital collections of instrument recordings. In this context, a piano sample library provides the full range of sounds produced by a piano. Creating a piano sample library is a time consuming and repetitive process, which demands a trained pianist to play the 88 keys of the piano at as many as 20 loudness levels. Additionally, it requires achieving the different types of sounds produced by the piano, such as sustain resonances, sympathetic resonances, release sounds, and key-bed noises. The goal in this thesis is to describe, design, develop, test and discuss a state-of-the-art system to precisely trigger piano keys and aid the construction of piano sample libraries. The system is capable of triggering piano keys at multiple loudness levels as well as estimating temporal start and end markers used for sample cropping. Moreover, the system allows for achieving the multiple types of sounds produced by the piano. It focuses on acoustic pianos, but can potentially be used with other keyboard-based instruments, such as synthesizers and electric pianos. The detailed system is non-invasive, i.e. no sensors or actuators are inserted into the instrument mechanics. In short, a linear actuator is vertically positioned above the piano key with a noise-protected mounting rack. The position of the actuator slider is controlled by an embedded proportional–integral–derivative (PID) controller. In an upper level, a control software analyses the sounds captured by a microphone, both in the spectral and temporal domains. Finally, the system executes movements that mimic the finger of a trained pianist, meeting constraints in position, speed, acceleration and time.