BaroPoser: Real-time Human Motion Tracking from IMUs and Barometers In Everyday Devices > 자유게시판

• 목록
• 답변
• 글쓰기
• 게시판 리스트 옵션
  • 수정
  • 삭제

In recent times, tracking human movement utilizing IMUs from everyday devices comparable to smartphones and iTagPro portable smartwatches has gained rising recognition. However, because of the sparsity of sensor measurements and the lack of datasets capturing human movement over uneven terrain, current strategies typically wrestle with pose estimation accuracy and are typically limited to recovering movements on flat terrain solely. To this finish, we current BaroPoser, the first method that combines IMU and barometric information recorded by a smartphone and a smartwatch to estimate human pose and international translation in real time. By leveraging barometric readings, we estimate sensor peak changes, which provide precious cues for both improving the accuracy of human pose estimation and predicting global translation on non-flat terrain. Furthermore, we propose a local thigh coordinate body to disentangle local and global motion input for better pose illustration learning. We evaluate our methodology on each public benchmark datasets and actual-world recordings. Quantitative and iTagPro USA qualitative results demonstrate that our approach outperforms the state-of-the-art (SOTA) strategies that use IMUs solely with the same hardware configuration.

Human movement capture (MoCap) is an extended-standing and challenging drawback in laptop graphics and vision. It goals to reconstruct 3D human physique movements, with many functions in film production, gaming, iTagPro portable and AR/VR. Although imaginative and prescient-based mostly methods (Peng et al., 2021; Lin et al., 2024; Wang et al., 2024; Xiu et al., iTagPro smart tracker 2024) have made vital progress in this subject, they at all times require digital camera visibility and are sensitive to occlusions and lighting conditions. Some works have focused on monitoring human motion through accelerations and rotations recorded by body-worn Inertial Measurement Units (IMUs), which overcome the aforementioned limitations. Commercial solutions in this class require 17 or extra IMUs, which will be intrusive and time-consuming for utilization. Recent studies (Huang et al., 2018; Yi et al., 2021, 2022; Jiang et al., 2022; Van Wouwe et al., 2023; Yi et al., iTagPro portable 2024; Armani et al., 2024) have decreased the number of IMUs to six or fewer, hanging a stability between accuracy and practicality.

However, these strategies nonetheless require specialized IMU sensors, limiting their application in on a regular basis life. To sort out this problem, some research (Mollyn et al., 2023; Xu et al., 2024) leverage the IMUs already available in on a regular basis units (akin to telephones, watches, wristbands, and iTagPro portable earbuds) for human movement capture. These methods outline a set of typical system placement places like the pinnacle, wrists, or iTagPro portable pockets, and use up to 3 of them to estimate human pose and iTagPro website world translation via neural networks. However, the accuracy of their movement estimation remains restricted, as the issue is inherently beneath-constrained attributable to sparse and noisy IMU measurements out there in on a regular basis settings. This makes it tough to precisely recover both native body poses or world translations. On this paper, we current BaroPoser, the primary strategy that fuses IMU and barometric data from one smartwatch (worn on one wrist) and track lost luggage one smartphone (positioned in the thigh pocket of the alternative facet) to estimate full-body motions in actual time.

Along with IMU data, which has been widely used for MoCap, we propose to incorporate barometric readings from constructed-in sensors in on a regular basis devices reminiscent of smartphones and smartwatches. These readings provide details about absolute altitude, offering an additional function to enhance the accuracy of both pose and world translation estimation. Such vertical awareness is particularly invaluable in purposes comparable to AR/VR and fitness monitoring, the place altitude-delicate actions like stair climbing, squats, and jumps are frequent. Moreover, to raised exploit human motion priors in this setting with solely two sensors, we introduce a thigh coordinate system to symbolize local physique poses, which helps to decouple the native and international motions. Specifically, we outline a local coordinate frame for the sensor on the thigh and iTagPro bluetooth tracker treat it as the basis coordinate frame of the human local pose. Then, iTagPro portable both the input and output of the pose estimation community are represented on this root coordinate body. On this method, the worldwide and local motion info recorded by the sensors is disentangled naturally because the thigh sensor information the worldwide movement info while the wrist sensor records the native one.