Firefly algorithm and DNN for improved contactless heart rate measurement from videos

Area of Science:

• Biomedical Engineering
• Computer Science
• Artificial Intelligence

Background:

• Contactless heart rate monitoring is crucial for non-invasive health assessment.
• Existing methods for facial-based heart rate detection face challenges in accuracy and feature selection.
• Deep learning approaches offer potential for improved physiological signal analysis.

Purpose of the Study:

• To develop a novel, accurate, and non-invasive system for heart rate measurement using facial video analysis.
• To enhance feature selection and classification accuracy for heart rate detection.
• To integrate Firefly optimization with a Deep Neural Network (DNN) for superior performance.

Main Methods:

• Utilized facial video analysis for contactless heart rate measurement.
• Implemented a Deep Neural Network (DNN) architecture.
• Employed Firefly optimization for feature selection and model optimization.
• Evaluated performance using precision, recall, F-measure, and accuracy metrics.

Main Results:

• Achieved a precision score of 90.22% and a recall score of 94.46%.
• Outperformed existing state-of-the-art methods by up to 2.1% in precision, 1.5% in recall, and 2.2% in F-measure.
• Demonstrated superior performance in classifying clinically meaningful heart rate categories.

Conclusions:

• The proposed Firefly optimization-integrated DNN offers a dependable and accurate technique for non-contact heart rate measurement.
• This advancement holds significant promise for applications in healthcare, fitness tracking, and stress management.
• The method provides a clear advantage for non-contact heart rate classification compared to recent methodologies.