Musical Instrument Identification based on a Short-Time Spectral Analysis

Abstract:- In this paper, an implementation of an audio recognition system for personal computers is presented, combining methodologies of digital signal processing, machine perception, and statistical decision models. In particular, attention was given to musical tones, harmonics, and MIDI notes, that build the musical context to identify two musical instruments from their corresponding musical notes. Altogether nine musical instruments from electronic and acoustic devices have been sampled, with an 89 % correct performance for single-mode identification and 67% correct performance for double-mode identification. Key-Words:- pattern recognition, Bayes decision theory, statistical classifiers, musical analysis, harmonics