Frog Identification System Based on Local Means K-Nearest Neighbors with Fuzzy Distance Weighting

Abstract

Frog identification based on the vocalization becomes important for biological research and environmental monitoring. As a result, different types of feature extractions and classifiers have been employed. Yet, the k-nearest neighbor (kNN) is one of the popular classifiers and has been applied in various applications. This paper proposes an improvement of kNN in order to evaluate the accuracy of frog sound identification. The recorded sounds of 12 frog species obtained in Malaysia forest have been segmented using short time energy and short time average zero crossing rate while the features are extracted by mel frequency cepstrum coefficient. Finally, a proposed classifier based on local means kNN and fuzzy distance weighting have been employed to identify the frog species. Comparison of the system performances based on kNN, local means kNN and the proposed classifier i.e. fuzzy kNN with manual segmentation and automatic segmentation is evaluated. The results show the proposed classifier outperforms the baseline classifier with accuracy of 94.67 % and 98.33 % for manual and automatic segmentation, respectively.