A New Approach to Fall Detection Based on Improved Dual Parallel Channels Convolutional Neural Network.

Area of Science:

• Biomedical Engineering
• Gerontology
• Machine Learning

Background:

• Falls represent a significant cause of injury and mortality in individuals over 65.
• Effective fall detection systems are crucial for timely intervention and prevention of severe consequences.

Purpose of the Study:

• To develop and evaluate an advanced fall detection method using surface electromyography (sEMG) signals.
• To investigate the efficacy of an improved dual parallel channels convolutional neural network (IDPC-CNN) for classifying falls versus daily activities.

Main Methods:

• Comparison of time-domain features and spectrograms of sEMG signals using LDA, KNN, and SVM classifiers.
• Selection of spectrogram features as input for the proposed IDPC-CNN model.
• Performance evaluation of IDPC-CNN against SVM and other CNN architectures.

Main Results:

• Spectrogram features demonstrated superior pattern extraction and classification performance compared to time-domain features.
• The IDPC-CNN model achieved 92.55% accuracy, 95.71% sensitivity, and 91.7% specificity.
• IDPC-CNN outperformed comparative models in accuracy, efficiency, training, and generalization.

Conclusions:

• The proposed IDPC-CNN model effectively utilizes sEMG spectrogram features for accurate fall detection.
• This method offers a promising advancement in fall detection technology for the elderly population.
• IDPC-CNN presents a more effective and generalizable approach compared to existing methods.