DensePose from WiFi:Seeing Human Pose Through Walls

Summarizes research that uses wifi to do human pose estimation

1. DensePose from WiFi Seeing Human Pose Through Walls Imagine seeing human movement and posture, even through solid walls, using nothing but ordinary WiFi signals. This isn't science fiction. Carnegie Mellon University has pioneered a revolutionary wireless sensing technology that accurately reconstructs human body posture—without the need for cameras.

2. Presentation Overview This presentation will guide you through the innovative world of DensePose from WiFi, a technology that allows us to see human posture through walls using wireless signals. Introduction Scientists Behind the Innovation Technology Overview How It Works: The Mechanism Operational Workflow Recent Advances & Milestones Real-World Applications Privacy Considerations Future Impact & Potential Conclusion & Key Takeaways

3. Meet the Scientists Behind the Breakthrough Jiaqi Geng Jiaqi Geng is a distinguished researcher specializing in computer vision and artificial intelligence, particularly in areas concerning human-computer interaction and perception. His academic journey and early research at Carnegie Mellon University laid the groundwork for innovative approaches to understanding human motion and form. Key Contributions to DensePose from WiFi: Pioneering the integration of deep learning techniques for robust pose estimation from unconventional data sources. Developing advanced neural network architectures capable of interpreting complex WiFi signal patterns. Focusing on the precise mapping of dense human pose information, crucial for the project's success. Research Focus: Deep Learning, Computer Vision, Human Pose Estimation, Generative Models, and AI-driven perceptual systems. Dong Huang Dong Huang brings extensive expertise in wireless sensing systems and signal processing, with a profound understanding of how to extract meaningful data from ambient radio frequency environments. His work has consistently pushed the boundaries of non-invasive sensing technologies. Key Contributions to DensePose from WiFi: Designing the fundamental wireless sensing methodologies to capture nuanced WiFi signal perturbations caused by human movement. Developing algorithms for effective noise reduction and signal feature extraction from raw WiFi data. Bridging the gap between raw wireless measurements and interpretable data for computer vision models. Research Focus: Wireless Communication, Radar Systems, Signal Processing, Human Sensing, and IoT applications. Fernando De la Torre Professor Fernando De la Torre is a renowned expert in machine learning, computer vision, and signal processing, with a strong background in developing probabilistic models for analyzing human behavior. His leadership and theoretical depth have been instrumental in many groundbreaking projects. Key Contributions to DensePose from WiFi: Providing theoretical foundations and advanced machine learning models for the project, particularly in probabilistic inference. Guiding the architectural design of the overall system, ensuring robustness and accuracy in diverse environments. Leveraging his extensive experience in human motion analysis to validate and refine the pose estimation outputs. Research Focus: Machine Learning, Computer Vision, Affective Computing, Probabilistic Graphical Models, and Human Behavior Analysis. This interdisciplinary team from Carnegie Mellon University pioneered a revolutionary approach, transforming WiFi signals into detailed human pose maps through advanced neural networks and computer vision. Their combined expertise in deep learning, wireless sensing, and machine learning was uniquely positioned to tackle the complex challenge of "seeing" through walls, establishing a new paradigm in non-intrusive human sensing.

4. What is DensePose from WiFi? Core Innovation AI system reconstructing human body posture using WiFi signals Standard routers—no specialised hardware Works through walls, furniture, and in complete darkness

5. How Does It Work? The Core Technology 01 Signal Propagation WiFi signals bounce off human bodies, carrying detailed shape and motion data encoded in their electromagnetic waves. 02 CSI Analysis Channel State Information (CSI) captures signal phase and amplitude changes, revealing body movements and positioning in space. 03 Neural Network Processing Deep learning models analyse CSI data, learning patterns that correspond to specific body positions and movements. 04 DensePose Mapping Trained DensePose model maps signals to 24 distinct body regions in precise 3D coordinates, creating detailed pose estimation.

6. Workflow: From Signals to Pose Estimation DensePose NN Signal Processing CSI Collection WiFi Routers The complete pipeline transforms raw wireless signals into actionable human pose data through advanced signal processing and machine learning, enabling real-time multi-person tracking without cameras.

7. Recent Advances: Person-in-WiFi 3D (CVPR 2024) Multi-Person Detection Simultaneously tracks multiple people in the same environment using distributed WiFi devices. Transformer Architecture Employs state-of-the-art transformer-based end-to-end models for faster processing and improved accuracy. Precision Performance Achieves joint localisation errors of 9–12 centimetres, rivalling camera systems and LiDAR technology.

8. Real-World Application: Mrs. Eleanor's Smart Home Imagine Mrs. Eleanor, living independently. The Person-in-WiFi 3D technology transforms her home into a haven of discreet safety and peace of mind for her and her family. Discreet, Whole-Home Monitoring Small WiFi sensors are strategically placed in Mrs. Eleanor's home. These non-invasive devices continuously monitor her presence and movements in every room, even through walls and furniture, without any cameras or wearables. Instant Fall Detection If Mrs. Eleanor experiences a sudden change in posture or a fall, the system instantly detects it. Its advanced algorithms differentiate normal movements from critical events, ensuring high accuracy. Immediate Caregiver Alert Upon detecting a fall, the system automatically sends an immediate alert to her daughter, Sarah, who is registered as her primary caregiver. Sarah receives a notification on her phone, along with Mrs. Eleanor's last known location. Peace of Mind & Independence This allows Sarah to check in or send help quickly, ensuring Mrs. Eleanor's safety and well-being. Mrs. Eleanor maintains her independence, knowing she's protected, while Sarah gains peace of mind, all without intrusive surveillance.

9. The Privacy Paradox: Innovation vs. Invisible Surveillance ✓ Privacy Advantage No visual images captured—far less intrusive than cameras, protecting dignity and identity. ⚠ Covert Risk WiFi networks exist everywhere; signals could be intercepted or repurposed without user knowledge or consent. This technology raises urgent ethical questions: Can invisible sensing be truly consensual? Who controls WiFi data? How do we balance life-saving monitoring with the right to privacy in our homes and workplaces?

10. Why This Matters: Transforming Human Sensing 1 Cost Revolution Dramatically more affordable than LiDAR, radar, or camera systems—enabling widespread deployment in resource-constrained environments. 2 New Possibilities Unlocks applications in healthcare monitoring, workplace safety, smart buildings, robotics, and human–computer interaction. 3 Privacy-Preserving Path Opens the door to AI-powered monitoring systems that protect privacy whilst delivering life-saving and efficiency gains.

11. DensePose from WiFi: Key Takeaways & The Path Forward DensePose from WiFi represents a pivotal moment: This innovative technology transforms everyday wireless networks into powerful, invisible sensing tools, blending AI ingenuity with pervasive connectivity. Yet, with great capability comes profound responsibility and critical considerations for its future. Transformative Potential DensePose from WiFi offers transformative potential across healthcare, safety, and accessibility, promising to save lives and significantly enhance quality of life through innovative, invisible sensing applications. Responsible Development Realizing this potential responsibly requires immediate focus on establishing robust ethical frameworks, clear regulatory standards, and transparent user consent protocols to ensure public trust and safeguard privacy. A Human-Centric Future Our vision is a future where WiFi intelligently connects and understands us, built upon foundations of responsibility, transparency, and human-centric design, making pervasive sensing a powerful force for good in the world.