On the Comparison of Wearable Sensor Data Fusion to a Single Sensor Machine Learning Technique in Fall Detection

Abstract

In the context of the ageing global population, researchers and scientists have tried to find solutions to many challenges faced by older people. Falls, the leading cause of injury among elderly, are usually severe enough to require immediate medical attention; thus, their detection is of primary importance. To this effect, many fall detection systems that utilize wearable and ambient sensors have been proposed. In this study, we compare three newly proposed data fusion schemes that have been applied in human activity recognition and fall detection. Furthermore, these algorithms are compared to our recent work regarding fall detection in which only one type of sensor is used. The results show that fusion algorithms differ in their performance, whereas a machine learning strategy should be preferred. In conclusion, the methods presented and the comparison of their performance provide useful insights into the problem of fall detection.

Keywords: accelerometer; data fusion; fall detection; gyroscope; mHealth; smartphone; wearable sensors.

Figure 1

Schematic diagrams showing fusion architectures at data (a), feature (b), and decision (c) levels.

Figure 2

Block diagram of the complementary filter. The complementary filter removes noise (low-pass) from the accelerometer data and eliminates drift (high-pass) of gyroscope data.

Figure 3

Typical waveform of fall event from accelerometer sensor.

Figure 4

Block diagram of threshold selection procedure. (*) The best thresholds are the ones that yield a performance close to (0, 1) in the ROC curve.

Figure 5

Cumulative confusion matrix of Alg. 4 with data from the MobiAct dataset. For the wrong predictions the percentage of each ADL (CSI: Step-into car, CSO: Step-out car, STN: Go downstairs, STU: Go upstairs) and fall (BSC: Back-sitting-chair; FKL: Front-knees-lying; FOL: Forward-lying; SDL: Sideward-lying) type is reported.

Figure 6

Nemenyi post-hoc tests for the performance metrics: (a) sensitivity; (b) specificity; (c) precision; and (d) F1. The red horizontal line in the sensitivity metric test denotes that the Alg. 1 and Alg. 4 are not different.