Bird Sounds Classification Using Linear Discriminant Analysis

Visually identifying birds is a challenging task due to its mobile airborne nature especially in the dense jungle of the Borneo rainforest. Instead, bird sounds may be used for bird identification purpose. The goal of this study is to classify and identify Borneo regional birds from their sounds, by using simple classification method. Five (5) local bird species have been used to demonstrate the performance of the proposed method. Using audio signal processing techniques, thirty-five (35) features have been extracted from the bird sounds, which are then reduced using Linear Discriminant Analysis (LDA), before feeding the data to Nearest Centroid (NC) classifier. It has been demonstrated that the proposed method outperforms other complex methods such as support vector machine and K-nearest neighbor, with a prediction accuracy level of 96%.