Dual-modal approach combines muscle activity and first-person vision to capture richer human demonstration data for humanoid robots
AUSTIN, Texas, Sept. 01, 2026 (GLOBE NEWSWIRE) -- Wetour Robotics Limited (NASDAQ: WETO) ("Wetour Robotics" or the "Company"), a Physical AI infrastructure and wearable robotics company, today released a development demonstration of Orchestra combining surface electromyography (sEMG) with first-person vision to capture richer human-hand data for robot learning.
The system is designed to address two limitations of vision-only capture: cameras cannot directly observe how much force a hand applies, and they lose hand data when the hand is hidden behind an object or leaves the field of view. Orchestra combines muscle activity with visual position so the two signals can complement each other at the edge.
"Physical AI needs more than video. A camera can show where a hand moved, but not how hard it worked, and it can go blind at the exact moment contact happens. Orchestra is designed to add force-related information and continuity to human demonstration data," said Nan Zheng, Chief Executive Officer of Wetour Robotics.
Two Signals That Fill Each Other's Blind Spots
Orchestra combines Conductor, an 8-channel sEMG wristband, with VisionLink, a first-person camera. Vision provides spatial position and scene context. sEMG captures muscle activation, remains available during visual occlusion and may provide a physical signal before visible movement begins. The development architecture is designed to fuse both streams locally into one synchronized record of movement, effort and action timing.
In a representative internal carrying task, the vision-only pipeline failed to locate the hand in 21.8% of frames, including a longest continuous dropout of 4.32 seconds. The architecture is designed to use sEMG as a complementary signal during those missing visual intervals. The cross-modal correction strategy remains in validation.
What the Demonstration Shows
- Continuous digital hand. The current on-device model outputs 20 joint angles rather than only classifying a small set of preset gestures. In an internal model benchmark, processing a one-second sEMG window required 50.4 milliseconds with zero lookahead. This measurement reflects model processing time, not end-to-end system latency.
- Force as a measurable dimension. The force-estimation pipeline is designed to convert sEMG into scale-calibrated, kilogram-equivalent grasp-force estimates. Calibration uses a scale as the reference and fits the response to the wearer and wearing session. The pipeline has been validated end-to-end on synthetic data; validation with live wristband force recordings remains under development.
- Action-ready data. The system is being developed to layer events such as reach, grasp, hold and release over continuous hand motion, creating structured demonstrations for imitation learning and robot training.
Five real-world tasks. The demo shows packing a lunch box, sorting pills, disassembling a pen, measuring a drone with calipers and installing a drone propeller - tasks combining precision, changing hand effort and frequent visual occlusion.
Why It Matters for Humanoid and Embodied AI
Human demonstrations are becoming an important input for training robots in real environments. Wetour Robotics believes adding force-related data and greater resilience to visual occlusion can make that data more useful for fine manipulation, compliant control and future human-robot collaboration.
Demonstration videos are available at www.wetourrobotics.com and on the Company’s LinkedIn and X channels under Wetour Robotics and @WETO_IR_TEAM.
About Wetour Robotics Limited
Wetour Robotics Limited (NASDAQ: WETO) is a Physical AI infrastructure and wearable robotics company headquartered in Austin, Texas. The Company is developing Orchestra, a Physical AI platform that connects intelligent agents to the physical world through wearable sensors, visual intelligence, edge AI computing and connected machines. Conductor is the Company’s wrist-worn neuromuscular interface, and VisionLink is its visual intelligence module.
Forward-Looking Statements
This press release contains forward-looking statements regarding the development, integration, validation, performance, applications and commercialization of Orchestra, Conductor, VisionLink, sEMG-vision fusion, force estimation, multimodal data capture and humanoid or embodied-AI applications. These statements are subject to risks including technology-development and integration risk, differences between synthetic and real-world data, calibration and accuracy limitations, competition, customer acceptance, capital requirements and other risks described in the Company's filings with the U.S. Securities and Exchange Commission. Actual results may differ materially. The Company undertakes no obligation to update forward-looking statements except as required by law.
Investor Relations Contact
Annabelle Li
Head of Investor Relations
Wetour Robotics Limited
ir@wetourrobotics.com
