Automatic Music Transcription Software

FPGA-Based Automatic Music Transcription Software

Presenter: Brandon Fong

Presenter Status: Graduate student

Academic Year: 20-21

Semester: Spring

Faculty Mentor: Nansong Wu

Department: Engineering

Screenshot URL: https://drive.google.com/uc?id=1mO-HrUEOCmTNv5Y61moO3xzI5-Reqxld

Abstract:
Field Programmable Gate Arrays (FPGA) are very popular for their hardware acceleration.  FPGAs have become a source of interest to many industry vendors like Microsoft. In this research, we examine the advantages software applications has while using FPGAs for complex computations.  This thesis will realize an application, Automatic Music Transcription (AMT) in real time, that will run on a Pynq Z2 FPGA board.  AMT is a complex procedure that analyzes audio and transcribes it, producing sheet music.  There are existing applications for AMT, some of which require processing time before presenting results.  The requirement for this application is to provide music transcriptions while taking advantage of the on board programmable logic, therefore accelerating the transcription process.  The main approach for this thesis is to take advantage of Pynq’s hardware libraries that allow the user to exploit Programmable logic from a Linux kernel, through a Python interface.