Elephant sound classification using machine learning algorithms for mitigation strategy
T. Thomas Leonid, R. Jayaparvathy · International Journal of Engineering Systems Modelling and Simulation · 2024
Conflicts between humans and elephants have become a wide problem in the agricultural and forest sectors, posing a threat to human lives and inflicting significant resource loss. This paper presents and compares the results of feature extraction techniques for detecting elephant voice signals. Support vector machine (SVM) classifiers, K-nearest neighbour (KNN) classifiers, Naive Bayes classifiers and convolutional neural network (CNN) classifiers all use the recovered features as inputs. The performance of all feature extraction techniques are validated and compared on elephant voice signals. The experimental results have confirmed that highest testing classification accuracy of 84% is resulted from CNN classifier with discriminatory features from the voice. This signifies that the different techniques of feature extraction technique have immense potential than other techniques in identifying elephant voice signal.