Multiphase Identification Algorithm for Fall Recording Systems Using a Single Wearable Inertial Sensor.

Area of Science:

• Biomechanics
• Machine Learning
• Gerontology

Background:

• Fall-related information is crucial for clinical diagnoses and developing effective fall prevention strategies.
• Understanding different fall phases, including falling time and landing responses, provides vital data for clinical assessment.
• Current systems may lack the granularity to capture these distinct fall phases accurately.

Purpose of the Study:

• To propose an automatic multiphase identification algorithm for phase-aware fall recording systems.
• To enable fine-grained data collection on various fall phases for clinical applications.
• To enhance the diagnostic and preventative capabilities of fall monitoring systems.

Main Methods:

• A pilot study involving seven young adults performing fall experiments.
• Utilizing a single inertial sensor worn on the waist to collect body movement data (525 trials).
• Combining machine learning techniques (SVM, kNN, Naïve Bayesian, Decision Tree, Adaptive Boosting) with a fragment modification algorithm to identify five fall phases: pre-fall, free-fall, impact, resting, and recovery.

Main Results:

• The proposed multiphase identification algorithm, particularly using the k-Nearest Neighbors (kNN) technique, demonstrated strong performance.
• Achieved 82.17% sensitivity, 85.74% precision, 73.51% Jaccard coefficient, and 90.28% accuracy in identifying fall phases.
• The fragment modification algorithm effectively detected data fragments inconsistent with neighboring data points.

Conclusions:

• The developed automatic multiphase identification algorithm shows significant potential for clinical use.
• It can provide automatic, fine-grained fall information for improved clinical measurement and assessment.
• This technology can aid healthcare professionals in making more accurate diagnoses and planning personalized fall prevention strategies.