Variational Mode Decomposition of Seismic Signals for Detection of Moving Elephants

Elephant monitoring in the vicinity of railway tracks is an important area of research for the reduction of rail-induced elephant accidents. Due to the lack of an autonomous system, effective detection and classification of pachyderms’ movement remain a challenge for early warning applications. This article presents a hybrid approach of using variational mode decomposition (VMD) coupled with feature extraction, to classify elephants’ movement in a forest environment. Furthermore, temporal, spectral, and cepstral domain features are extracted from the principal modes of VMD and are used to classify the seismic signatures of other movements, such as deer, humans’ movement, trains, and electrical noise using support vector machines (SVMs). The proposed method is compared with the traditional signal decomposition approaches, such as empirical mode decomposition (EMD) and empirical wavelet transform (EWT). The classification results for the elephants’ class show an average improvement of ~23% and ~16% in F1 score and false-negative rate, respectively, in comparison with the EMD and EWT. The VMD of the seismic signals has an average accuracy of ~73% ± 5% for the classifier SVMs with quadratic kernel.