CardiMamba: AI-Driven Remote Heart Rate Estimation

The CardiMamba model integrates RGB video and radar RF data for remote heart rate estimation, overcoming the limitations of traditional contact-based methods and single-mode signals. It consists of three core components: Dual-level Feature Extraction and Alignment, Bidirectional Feature Interaction, and Bidirectional Feature Fusion. The model extracts features from both RGB and RF data, merges them in a shared space, and fuses them to generate new feature representations. This fusion enhances accuracy and robustness by combining visual information from RGB with physiological signals from RF. The CardiMamba model offers a more accurate and stable heart rate estimation method. This work is 10 months of Zheng’s hard work and has been proved by complete experiments, including comparison experiments with previous methods and ablation experiments. This work is about to be published in an international journal. My research exhibition offers an interactive experience where visitors can witness my technology firsthand. Using a camera and radar to monitor heart rate in real-time, visitors will see how my multimodal fusion approach works. This live demonstration showcases non-contact heart rate estimation, combining RGB video data and radar RF signals for accurate, reliable health monitoring without physical sensors. Visitors will appreciate the practical implications in healthcare, fitness, and security. The combination of advanced deep learning and sensor fusion will intrigue attendees, highlighting the technology's potential to revolutionize personal health monitoring and broader medical advancements.