Hi-Fi audio and machine learning implementation in MIDI-controller prototype development

Abstract: Musical Instrument Digital Interface(MIDI) is a technical standard that specifies acommunication protocol for digital instrument data and is mainly used by computers,synthesizers and electronic keyboards. A MIDI controller is any hardware or software that cangenerate and transfer MIDI data to a MIDI device, usually to play a sound or change controlparameters within a music software. An important property of a MIDI controller is to intuitively go from idea to execution withminimal effort. In this respect machine learning has really opened up for new musicalexpressions within musical devices. During this project a prototype with an implementedmachine learning algorithm which could deliver a wider array of functionality with a simplerinterface. The device developed should also be able to play Hi-Fi audio with at least 96 kHzsampling frequency and 24-bit bit-resolution. Without professional measurement equipment audio quality could not be completelymeasured. However, the development of the completed prototype as a whole was successful. Drawn symbols could be classified with a 70% accuracy, Hi-Fi audio could be played and MIDI-messages could be sent over USB to a music software.