Acceleration Magnitude at Impact Following Loss of Balance Can Be Estimated Using Deep Learning Model

Area of Science:

• Biomechanics
• Wearable Technology
• Artificial Intelligence

Background:

• Pre-impact fall detection aims to prevent injuries by predicting falls before ground contact.
• Impact acceleration magnitude is a critical factor in injury severity and a key parameter for protective device design.

Purpose of the Study:

• To propose a novel method for predicting impact acceleration magnitude using a single inertial measurement unit (IMU) sensor and a deep learning model.
• To evaluate the effectiveness of data augmentation techniques in improving prediction accuracy.

Main Methods:

• A bi-directional long short-term memory (LSTM) regression model was developed to predict impact acceleration magnitude.
• Experiments involved 24 healthy participants wearing a single IMU sensor at the waist, collecting tri-axial accelerometer and angular velocity data during simulated falls.
• Data augmentation techniques were applied to increase the dataset size and improve model performance.

Main Results:

• The proposed model achieved a mean absolute percentage error (MAPE) of 6.69 ± 0.33% and an r-value of 0.93 with data augmentation.
• A 4-fold increase in training data resulted in a significant 45.2% reduction in MAPE.
• The study demonstrates the feasibility of using predicted impact acceleration for real-time fall prevention systems.

Conclusions:

• Impact acceleration magnitude can be accurately predicted pre-impact using IMU data and deep learning.
• Data augmentation significantly enhances the performance of fall impact prediction models.
• This predictive capability is crucial for optimizing the deployment of wearable protective devices, such as airbag systems, to minimize fall-related injuries.