Title: A comparative study of the difference between MFCC and PLP in the recognition of sound

Authors: Sabur Ajibola Alim; Nahrul Khair Alang Md. Rashid; Md. Mozasser Rahman

Addresses: Mechatronics Engineering Department, Kulliyah of Engineering, International Islamic University Malaysia (IIUM), 53100, Gombak, Malaysia ' Mechatronics Engineering Department, Kulliyah of Engineering, International Islamic University Malaysia (IIUM), 53100, Gombak, Malaysia ' Mechatronics Engineering Department, Kulliyah of Engineering, International Islamic University Malaysia (IIUM), 53100, Gombak, Malaysia

Abstract: Sound is one of the most important tools for classification, recognition and identification of objects in the environment. The raw sound signal is complex and is not suitable to be feed as input to the sound identification system; hence the need for a good front-end arises. The identification rate using the RNN classifier and MFCC is 72.7%, 73.7%, 78.9% 57.1% and 58.3% for aircraft, car, rain, thunder and train respectively as compared to what was obtained by using MLP. 31.6%, 19.4%, 18.5%, 38.0% and 26.4% decline is achieved for aircraft, car, rain, thunder and train respectively when comparing between MLP and RNN for MFCC. As far as sound recognition using the input used in this experiment is concerned, MFCC outperforms PLP and MFCC and PLP using MLP as classifier.

Keywords: Mel frequency cepstral coefficients; MFCC; perceptual linear prediction; PLP; multilayer perceptron; MLP; recurrent neural networks; RNNs; sound recognition.

DOI: 10.1504/IJMEI.2013.053331

International Journal of Medical Engineering and Informatics, 2013 Vol.5 No.2, pp.145 - 151

Published online: 28 Jan 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article