Human-elephant conflict (HEC) takes hundreds of human and elephant lives, every year. Though there are many techniques and systems deployed in African and Asian countries to mitigate this conflict, none of them have provided efficient solutions. The success of a modern HEC mitigating system heavily depends on its capability to detect the presence of an elephant. In most of the existing systems, their superior accuracy of detecting an elephant's presence is limited by certain conditions such as the ability to capture a good quality image containing the elephant. In this paper, we propose an alternative elephant detection system which uses odour of elephant urine and the earflap sound as detection parameters and also a support vector classification for decision making. The proposed system can produce higher accuracies in detecting wild elephants present in a 35m radius, under all conditions. Thus, the proposed system is much more superior over the existing elephant detection systems and will be an effective tool to mitigate the HEC.