Towards 3D human pose construction using wifi

Jiang, Wenjun; Xue, Hongfei; Miao, Chenglin; Wang, Shiyang; Lin, Sen; Tian, Chong; Murali, Srinivasan; Hu, Haowei; Sun, Zhi; Lu, Shan

doi:10.1145/3372224.3380900

Cited by 147 publications

(62 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this work, we aim to learn the features of a space using Wi-Fi CSI data which has substantially more information than RSS [22]. Modern Wi-Fi signals are transmitted through multiple subcarriers (e.g.…”

Section: B Wi-fi Csimentioning

confidence: 99%

Secure Region Detection Using Wi-Fi CSI and One-Class Classification

et al. 2021

View full text Add to dashboard Cite

Location-based authentication for Internet of Things (IoT) devices can be used to permit activation and participation only to those devices that are in their predefined areas. It secures the devices from being used elsewhere (e.g. misplacement or theft), and also secures the system by preventing nonauthorized devices from joining the network. However, absolute location is not required for such purposes; only whether the device is within the designated 'secure' region or not matters, which we define as the secure region detection problem. In this work, we propose SWORD, the first secure region detection scheme based on channel state information (CSI) of Wi-Fi and deep one-class classification (OCC) technique. OCC can be trained using data only from the inside of a secure region and no negative reference point is required, a critical advantage considering that outside of a secure region is practically unbounded. Our real-world experiment results show that SWORD can achieve 99.14% true-negative rate (TN, successfully rejecting devices not in secure region) and an acceptable true-positive (TP) of 76.90% for practical usage. Furthermore, there is an user-adjustable trade-off between TN and TP based on application requirement, and TP can be improved to 97.92% without a big loss of TN using simple automatic repeat mechanism.

show abstract

Section: B Wi-fi Csimentioning

confidence: 99%

Secure Region Detection Using Wi-Fi CSI and One-Class Classification

et al. 2021

View full text Add to dashboard Cite

show abstract

“…This is a main challenge for all WiFi-based approaches, as WiFi signals exhibit significantly different propagation patterns in different environments. To address this issue and achieve cross-environment generalization, the authors of WiPose [ 89 ] proposed to utilize 3D velocity profiles obtained from WiFi signals in order to separate posture-specific features from the static background objects. Their approach achieved an accuracy of up to 2.83 cm (mean per-joint position error), but is currently limited to a single non-moving person.…”

Section: Related Workmentioning

confidence: 99%

A Baseline for Cross-Database 3D Human Pose Estimation

Rapczyński

Werner

Handrich

et al. 2021

Sensors

View full text Add to dashboard Cite

Vision-based 3D human pose estimation approaches are typically evaluated on datasets that are limited in diversity regarding many factors, e.g., subjects, poses, cameras, and lighting. However, for real-life applications, it would be desirable to create systems that work under arbitrary conditions (“in-the-wild”). To advance towards this goal, we investigated the commonly used datasets HumanEva-I, Human3.6M, and Panoptic Studio, discussed their biases (that is, their limitations in diversity), and illustrated them in cross-database experiments (for which we used a surrogate for roughly estimating in-the-wild performance). For this purpose, we first harmonized the differing skeleton joint definitions of the datasets, reducing the biases and systematic test errors in cross-database experiments. We further proposed a scale normalization method that significantly improved generalization across camera viewpoints, subjects, and datasets. In additional experiments, we investigated the effect of using more or less cameras, training with multiple datasets, applying a proposed anatomy-based pose validation step, and using OpenPose as the basis for the 3D pose estimation. The experimental results showed the usefulness of the joint harmonization, of the scale normalization, and of augmenting virtual cameras to significantly improve cross-database and in-database generalization. At the same time, the experiments showed that there were dataset biases that could not be compensated and call for new datasets covering more diversity. We discussed our results and promising directions for future work.

show abstract

“…This metric can evaluate the overall performances of the location error, pose error, shape error, and gender error. Average Joint Localization Error (S) [16,48]. This metric is defined as the average Euclidean distance between the joint locations of the predicted human mesh and the ground truths for all the subjects and activities.…”

Section: Metricsmentioning

confidence: 99%

“…Recently, researchers have put significant efforts towards building intelligent wireless sensing systems, which aim to perceive and understand human activities by leveraging pervasive wireless signals. Thus far, the most remarkable achievement in this effort is the construction of human skeletons from the signals reflected off the human body [16,34,35,47,49]. Having the skeletal representations, a follow-up question arises: Is the information contained in the RF signal rich enough to further reconstruct the shape of the human body from which we can tell not only the height but also the somatotype, weight, and even the gender of the monitored subject?…”

Section: Introductionmentioning

confidence: 99%

mmMesh

Xue

Yan

Miao³

et al. 2021

Proceedings of the 19th Annual International Conference on Mobile Systems, Applications, and Services

Self Cite

View full text Add to dashboard Cite

In this paper, we present mmMesh, the first real-time 3D human mesh estimation system using commercial portable millimeterwave devices. mmMesh is built upon a novel deep learning framework that can dynamically locate the moving subject and capture his/her body shape and pose by analyzing the 3D point cloud generated from the mmWave signals that bounce off the human body. The proposed deep learning framework addresses a series of challenges. First, it encodes a 3D human body model, which enables mmMesh to estimate complex and realistic-looking 3D human meshes from sparse point clouds. Second, it can accurately align the 3D points with their corresponding body segments despite the influence of ambient points as well as the error-prone nature and the multi-path effect of the RF signals. Third, the proposed model can infer missing body parts from the information of the previous frames. Our evaluation results on a commercial mmWave sensing testbed show that our mmMesh system can accurately localize the vertices on the human mesh with an average error of 2.47 cm. The superior experimental results demonstrate the effectiveness of our proposed human mesh construction system. CCS CONCEPTS• Human-centered computing → Ubiquitous and mobile computing; • Computer systems organization → Embedded and cyber-physical systems.

show abstract

Towards 3D human pose construction using wifi

Cited by 147 publications

References 51 publications

Secure Region Detection Using Wi-Fi CSI and One-Class Classification

Secure Region Detection Using Wi-Fi CSI and One-Class Classification

A Baseline for Cross-Database 3D Human Pose Estimation

mmMesh

Contact Info

Product

Resources

About