October 2, 2025SensorsOpen Access

Fall Detection by Deep Learning-Based Bimodal Movement and Pose Sensing with Late Fusion

View Full Paper

Ask AI

Bookmark

Authors

HRHaythem RehoumaUniversité du Québec à Montréal MBMounir BoukadoumUniversité du Québec à Montréal

Discussion

Member takes

Implication

This framework combines inertial and visual data to improve fall detection accuracy in elderly individuals, suggesting significant potential for real-time applications.

Key Points

The proposed bimodal deep learning approach achieved an impressive F1-score of 97.3%, showcasing its effectiveness.
The false-positive rate was significantly reduced to 3.6%, far lower than the 11.3% from IMU-only and 8.9% from vision-only methods.
A custom dataset was used for experimental evaluation, capturing simulated falls and routine activities under various lighting conditions.
The approach employs a memory-based autoencoder for detecting movement abnormalities and an attention-based model for visual pose analysis.

Cite This Study

Rehouma et al. (2025) studied this question.

synapsesocial.com/papers/68de68ea83cbc991d0a213ce https://doi.org/10.3390/s25196035

View Full Paper

Ask AI

Bookmark

Implication

This framework combines inertial and visual data to improve fall detection accuracy in elderly individuals, suggesting significant potential for real-time applications.

Fall Detection by Deep Learning-Based Bimodal Movement and Pose Sensing with Late Fusion

Discussion

Member takes

Implication

Key Points

Cite This Study

Fall Detection by Deep Learning-Based Bimodal Movement and Pose Sensing with Late Fusion

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider