Katydids acoustic classification on verification approach based on MFCC and HMM

Abstract

This work presents a new proposal towards the development of an intelligent system for automatic classification of katydids. Katydid is the common name of a certain large, singing, winged insects that belongs to the long-horned grasshopper family (Tettigoniidae) in the order of the Opthoptera. We propose a sound parameterization using Mel Frequency Cepstral Coefficients (MFCC) because these coefficients approximate the human auditory system's response more closely than linear-spaced frequencies. This proposal is based on the use of a HMM classifier to process the MFCCs. Our proposal is based on two approaches, identification and verification; and it has obtained 99.31% of accuracy in the identification stage and has increased to 99.97% of accuracy in the verification stage.