What problem does this paper attempt to address?

This paper aims to solve the problem of safety monitoring of personnel operations in power stations. Traditional monitoring methods such as wearable devices and video surveillance have limitations such as high cost, dependence on light conditions, and visual blind areas. Therefore, the paper proposes a device - free human pose estimation framework based on Channel State Information (CSI) - PowerSkel, to address these challenges. PowerSkel uses self - developed CSI sensors to form a mutual - inductance network and constructs a CSI collection scheme specifically for power scenarios. This scheme significantly reduces the deployment cost and complexity, and at the same time, a sparse adaptive filtering algorithm is designed to pre - process CSI signal interference in the power environment. In addition, the paper proposes a knowledge distillation network (CKDformer) based on collaborative learning and self - attention mechanisms, which is used to extract features from CSI and establish the mapping relationship between CSI and key points. Experimental results show that PowerSkel performs excellently in real - world power stations and can achieve high - precision human pose estimation even in dark environments. ### Main Contributions 1. **Proposed a human pose estimation method in power stations**: Use CSI for human pose estimation in power scenarios and develop data processing methods suitable for power scenarios. 2. **Designed the PowerSkel system**: This is a novel CSI mutual - inductance solution, using self - developed CSI sensors, and is the first CSI - based human pose estimation system in power stations, which can still work effectively especially in power stations with limited lighting conditions. 3. **Proposed the CKDformer model**: This is a knowledge distillation network based on collaborative learning and self - attention mechanisms. Through knowledge sharing among multiple Conformers, the transfer of global and local information in CSI channels is achieved. Use the key - point coordinates obtained by Kinect for cross - modal supervision, thereby achieving human pose estimation in power scenarios. ### Method Overview - **Data collection and processing**: Deploy multiple CSI sensors and Kinect, and synchronously transmit data to the server through wireless communication. Use the sparse adaptive filtering algorithm (SAF) to pre - process CSI signals to enhance signal quality and reduce noise. - **Model training**: Use Conformer as the core architecture, and through the collaborative knowledge distillation (CKD) method, make multiple student models share channel features and relationships to optimize the model training process. - **Pose estimation**: Finally, achieve accurate human pose estimation using only CSI input. ### Technical Details - **Sparse adaptive filtering algorithm (SAF)**: Update the filter coefficients by constructing a dictionary and solving the sparse representation to achieve the reconstruction of CSI signals. - **CKDformer model**: Combine the local and global feature extraction capabilities of Conformer, and through the optimal transport (OT) loss function, achieve knowledge sharing among student models and improve the model's prediction performance. ### Experimental Results - **High - precision pose estimation**: Experimental results show that the PCK@50 of PowerSkel in real - world power stations reaches 96.27%, and can achieve significant visual effects even in dark environments. In conclusion, this paper proposes a low - cost, high - precision human pose estimation method that is not affected by lighting conditions, providing a new solution for safety monitoring in power stations.

PowerSkel: A Device-Free Framework Using CSI Signal for Human Skeleton Estimation in Power Station

PowerSkel: A Device-Free Framework Using CSI Signal for Human Skeleton Estimation in Power Station

Mpose: Environment- and Subject-Agnostic 3D Skeleton Posture Reconstruction Leveraging a Single Mmwave Device

Mmskeleton: 3D Human Skeleton Estimation Using Millimeter Wave Radar Sparse Point Clouds

Rapid: A Multimodal and Device-free Approach Using Noise Estimation for Robust Person Identification

Skeleton-based human activity recognition with wifi CSI using a hybrid approach combining convolutional neural network and long short term memory

Skeleton-Based Human Pose Recognition Using Channel State Information: A Survey

CSI-Former: Pay More Attention to Pose Estimation with WiFi

A novel approach for simultaneous human activity recognition and pose estimation via skeleton-based leveraging WiFi CSI with YOLOv8 and mediapipe frameworks

Seeing the Unseen: Wifi-based 2D Human Pose Estimation via an Evolving Attentive Spatial-Frequency Network

Design of a Wireless Multiple-Mode Human Behavior Evaluation System.

3D Human Pose Estimation for Free-from and Moving Activities Using WiFi

Towards 3D Human Pose Construction Using Wifi.

WiFi-Enabled Smart Human Dynamics Monitoring

MDPose: Human Skeletal Motion Reconstruction Using WiFi Micro-Doppler Signatures

Human posture estimation and action recognition on fitness behavior and fitness

Leveraging Digital Perceptual Technologies for Remote Perception and Analysis of Human Biomechanical Processes: A Contactless Approach for Workload and Joint Force Assessment

A Real-time Human Pose Estimation Approach for Optimal Sensor Placement in Sensor-based Human Activity Recognition

Accurate Human Pose Estimation using RF Signals

3D Human Pose Estimation for Free-form Activity Using WiFi Signals

Design of dynamic ski motion capture and human posture detection model based on spatio-temporal Transformer and wireless device tracking technology