PIPTO: Precise Inertial-Based Pipeline for Threshold-Based Fall Detection Using Three-Axis Accelerometers.

Area of Science:

• Biomedical Engineering
• Gerontology
• Computer Science

Background:

• Falls are the second leading cause of death after traffic incidents, disproportionately affecting the elderly.
• Existing fall detection methods utilize various sensors and techniques, including machine learning and heuristic-based approaches.
• Combining heuristic techniques with machine learning models can reduce computational costs in fall detection systems.

Purpose of the Study:

• To present a low-complexity pipeline for fall detection using a threshold-based technique.
• To develop a system adaptable to various acceleration sensors and data frequencies.
• To enable the detection of multiple falls within a time series and specify their temporal range.

Main Methods:

• A threshold-based technique is applied to data from a three-axis accelerometer.
• The system processes variable input lengths and detects falls based on the sum vector magnitude.
• The developed framework, PIPTO, is implemented in Python and C.

Main Results:

• High performance in fall detection was achieved on multiple datasets.
• Sensitivity reached 90.40% on MMsys and 91.56% on KFall.
• Specificity was recorded at 93.96% on MMsys and 85.90% on KFall.

Conclusions:

• The proposed PIPTO pipeline offers a computationally efficient and accurate solution for fall detection.
• The system's adaptability and open-source nature facilitate its adoption by the research community.
• PIPTO contributes to enhanced safety for the elderly by providing reliable fall detection capabilities.