Musical Instrument Recognition using Mel-Frequency Cepstral Coefficients and Learning Vector Quantization

International Conference on Informatics, Engineering, Science and Technology

计算机科学;工业技术

Maliki, I.^1 ; Sofiyanudin^1

Department of Informatic Engineering, Faculty of Engineering and Computer Science, Universitas Komputer Indonesia, Jln. Dipatiukur no 112-116, Bandung

40132, Indonesia^1

关键词: Learning Vector Quantization;    Mel frequency cepstral co-efficient;    Music information retrieval;    Musical instrument recognition;    Performance tests;    Sound source;    Subtask;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/407/1/012118/pdf
DOI  :  10.1088/1757-899X/407/1/012118

来源: IOP

【 摘 要 】

Musical instrument recognition is an essential subtask in many application regarding in music information retrieval. This research aims at extending the previous research saying that the MFCC used to feature extraction and LVQ method used to classification has good accuracy for musical instrument recognition. To test the methods described have been implemented in an android based application. Looking at the presented results, this research then focuses on to implementation that method for recognition musical instrument based on Aerophone, Electrophone, Chordophone, Idiophone, and Membranophone. The test was performed used 750 dataset with 5 sound source classes, the result of the performance test show that methods has 94.80% accuracy. It can be concluded that MFCC and LVQ methods can be implemented to recognize musical instruments.