Comparison of fall detection methods and the prospect of model-based: pattern recognition on green computing implementation by using smartphone

Abstract

Portable healthcare solution to detect fall is needed for special care people and elderly, especially if they want to live independently. If this solution needs special made device by the factory, it requires huge electric consumption, lots of materials, and can contribute to the environmental pollution. Smartphone is one of the most promising device to be used as fall detection because it has lot of potentials: easy to find, compact, has sensors, and can even be accepted by elderly. There are two main approaches in fall detection: threshold-based method and model-based: pattern recognition method. Threshold-based usually been used in real life because it can be developed quite quickly and use minimal computation process to provide movement classification result (fall or non-fall). Model-based: pattern recognition method (using machine learning) has relatively better accuracy result but requires a long training time, a lot of resources to process, and more difficult to be developed. The researchers present the comparison results of threshold-based and model-based: pattern recognition; and optimizing the model in machine learning so the model has the prospect to be implemented in applications to support green computing.