A feasibility study of using smartphone built-in accelerometers to detect fall portents

• We used a smartphone (built-in accelerometers) to detect falls and fall portents.
• We designed two experiments to evaluate accuracy of three threshold algorithms.
• The sensitivity and specificity were 100% and 93.6%–96.1% in detecting falls.
• The best performer had a satisfactory accuracy rate of 88.5% in detecting fall portents.
• Detecting falls and fall portents is feasible in a tiling scenario.

Fall accidents contribute to nearly half of all fatalities in the construction industry in Taiwan. Detecting fall portents using a smartphone, which many people carry daily, may help reduce fall accidents if the accuracy is acceptable. We designed two experiments with three algorithms to evaluate how well a smartphone can detect both falls and fall portents in a tiling operation scenario. The experiments show that work-related motions barely affected the detection of falls, and the result had a sensitivity and specificity of 100% and 96.1%, respectively. However, for detecting portents, the work-related motions had quite a large impact on the gyroscope-based algorithm, which demonstrated an accuracy rate of only 4.3%, but had only limited impact on the accelerometer-based algorithm, which still show acceptable accuracy rates of 73.5% and 88.5%. We conclude that using a smartphone to detect falls and portents in a construction site is feasible.