Implementation of Mel Frequency Cepstral Coefficient and Dynamic Time Warping For Bird Sound Classification

Abstract

Lovebird (Agapornis) is a type of bird that has become the belle of new pet birds lately. The interest of the hobbyist in this one song is because Lovebird has a unique chirp. For beginner lovebird fans, the lack of knowledge and experience about lovebird birds results in various cases of fraud in choosing a quality lovebird. They were disappointed expensive lovebirds that had been purchased but did not match what was expected.
Lovebird chirping voice recognition can be learned and recognized through the learning process of speaker recognition, which is part of voice recognition. Speaker recognition captures the frequency of the lovebird's voice, then compares it with the sound frequency of the existing training data. The sound frequency and the long duration of chirping of lovebird birds will be extracted through the Mel-Frequency Cepstral Coefficient (MFCC) method. Information in the form of Mel Frequency Cepstrum Coefficients from input data and training data is then compared to the Dynamic Time Warping method. The methodology used in this study uses the grapple method.
The results of this study were obtained an accuracy value of sound validation by 80%. It is hoped that with the capabilities of this system, it can help bird chirping lovers know the sound quality of lovebird birds that are good, moderate, and less. Also, it can help the jury of birds chirping, so that it can be used as an accurate standard in classifying lovebird sounds.