The IoT based embedded system for the detection and discrimination of animals to avoid human–wildlife conflict

Abstract

Human–wildlife conflict (HWC) is one of the major crises in the valparai region of Anamalai Tiger Reserves (ATR). It is essential to reduce the HWC to save people from the wildlife and also to protect wildlife. In this paper, we propose an automated unsupervised system for the identification and classification of animals from their acoustic signal. The environment sound signals are captured using a microphone and the audio is stored in a.wav file and is sent to a base station through a radio frequency (RF) network. This system is processed with three steps (i) from the received audio signal initially, animal identification is done by extracting features of an animal signal using Mel frequency cepstral coefficient (MFCC) and classification of animal is performed by radial basis function (RBF) neural network, (ii) age estimation (calf/adult) is performed by autocorrelation, (iii) elephant state of mind (SOM) is detected by extracting features of an elephant acoustic signal using gammatone frequency cepstral coefficient (GFCC) and classification of various sounds of elephant are performed by support vector machine (SVM). Based on this, an early warning message which contains an animal type, age (calf/adult), elephant SOM, global positioning system (GPS) tracks its location information and all this information will be transmitted via global system for mobile communication (GSM) to the forest authorities, local communities, radio station or local channels indicating that an animal movement is near to forest border areas. The results were fed into a separate web page using the internet of thing (IoT).