Fall is one of the major cause of deaths in elderly along with other chronic diseases in all over the world. Therefore, it is important to find a cost effective, non-intrusive and lightweight solution for early fall detection and prevention in elderly. Several fall detection systems have been proposed, using the different types of sensors and techniques. In this paper, a novel fall detection technique, using the wearable SHIMMER™ sensors, is proposed, which identifies the fall event, using Mahalanobis distance on real-time data. It is more robust than other conventional distance measure techniques, followed in existing fall detection systems. We first developed a real dataset that consists of three daily life activities, such as walking, sitting (on) and getting up (from) a chair, and standing still. These activities are the main cause of fall in elderlies. The proposed algorithm was tested and validated, to identify the fall event. It produced the promising results, which are comparable to the state-of-the-art fall detection techniques.