FALLDROID: AN AUTOMATED SMART PHONE BASED FALL DETECTION SYSTEM USING MULTIPLE KERNEL LEARNING

Abstract

Common fall occurrences in the elderly population pose dramatic challenges in public healthcare domain. Adoption of an efficient and yet highly reliable automatic fall detection system may not only mitigate the adverse effects of falls through immediate medical assistance, but also profoundly improve the functional ability and confidence level of elder people. This paper presents a pervasive fall detection system developed on smart phones (SPs) namely, FallDroid that exploits a two-step algorithm proposed to monitor and detect fall events using the embedded accelerometer signals. Comprising of the threshold based method (TBM) and multiple kernel learning support vector machine (MKL-SVM), the proposed algorithm uses novel techniques to effectively identify fall-like events (such as lying on a bed or sudden stop after running) and reduce false alarms. In addition to user convenience and low power consumption, experimental results reveal that the system detects falls with high accuracy (97:8% and 91:7%), sensitivity (99:5% and 95:8%), and specificity (95:2% and 88:0%) when placed around the waist and thigh, respectively. The system also achieves the lowest false alarm rate of 1 alarm per 59 hours of usage, which is best till date.

Existing System

Multi-layer perceptron (MLP) neural network for fall event detection

Proposed System

Accordingly, we present an automated high performance SP-based fall detection system focusing on practical issues such as user convenience and power consumption. The proposed standalone fall detector is developed as an Android app, namely FallDroid, which uses the accelerometer sensor embedded in SPs. The designed application provides an elder friendly GUI and supports the two most convenient SP carrying locations: waist (belt/pouch) and thigh (pant pocket). In comparison with ML techniques, the proposed two-step algorithm is shown to be more power-efficient. In the first step, a low computational cost approach based on TBM is used, followed by the pattern recognition technique, multiple kernel learning support vector machine (MKL-SVM) in the second step which is rarely invoked. The battery consumption was analysed and reported for different scenarios. The recorded data sets were acquired from human trials conducted systematically in both, laboratory and free living environments. Finally, we report the offline and online classification results on fall-like ADLs such as lying on the floor, sudden stop after walking, accidentally hitting the sensor etc. to demonstrate the better performance of the presented system.

Architecture

BLOCK DIAGRAM