Effect of Natural Background Noise and Man-Made Noise on Automated Frog Calls Identification System

Frog identification based on their calls becomes important for biological research and environmental monitoring. However, identifying particular frog calls becomes challenging particularly when the frog calls are interrupted with noises either in natural background noise or man-made noise. Hence, an automatic identification frog call system that robust in noisy environment has been proposed in this paper. Experimental studies of 675 audio obtained from 15 species of frogs in the Malaysian forest and recorded in an outdoor environment are used in this study. These audio data are then corrupted by 10dB and 5dB noise. A syllable segmentation technique i.e. short time energy (STE) and Short Time Average Zero Crossing Rate (STAZCR) and feature extraction, Mel-Frequency Cepstrum Coefficients (MFCC) are employed to segment the desired syllables and extract the segmented signal. Subsequently, the Local Mean k-Nearest Neighbor with Fuzzy Distance Weighting (LMkNN-FDW) are employed as a classifier in order to evaluate the performance of the identification system. The experimental results show both of natural background noise and man-made noise outperform by 95.2% and 88.27% in clean SNR, respectively.