Musical Instrument Classification System Using Mel-Frequency Cepstral Coefficients

PROJECT GOAL

In this project, I will design a musical instrument identification system. The goal is to identify musical instruments from individual notes. I am given a database of 8 different instruments, each containing 1000 notes (i.e. different pitches). Each audio sample is sampled at 44.1KHz and is an isolated note of an instrument that varies in length depending on the instrument. My goal is to: (1) analyze the recordings in order to extract features that transform the raw waveform signals into a more compact and effective representation, (2) train an instrument classification model which would distinguish notes from different instruments.

DOCUMENT

PDF: pdf_icon

 

Examples

example1

example2

Advertisements