Wetour Robotics (NASDAQ: WETO) Demonstrates Conductor Neural Wristband with Training Powered by Meta’s Open emg2pose Dataset to Advance Physical AI Human-Machine Interaction and future physical-world models

About this update from Wetour Robotics Limited

Live demo turns wrist muscle signals into real-time 3D hand digital twins and gesture-to-text commands — creating an on-device human-intent data layer for robotics AUSTIN, Texas, June 30, 2026 (GLOBE NEWSWIRE) -- Wetour Robotics Ltd. (NASDAQ: WETO) ("Wetour" or the "Company"), a Physical AI and wearable-robotics infrastructure company, today released a new demonstration of Conductor, the sEMG (surface electromyography) neural wristband at the core of its Orchestra platform. In the demonstration, Conductor decodes muscle signals from an 8-channel wrist sensor into a real-time 3D hand pose — a live digital twin of the wearer's hand, rendered with no cameras and no gloves. Gestures into text, in real time. The demonstration also shows Conductor recognizing discrete hand gestures and converting them into text commands on screen in real time — turning deliberate gestures into typed input with no keyboard or touchscreen. Demonstration videos are available at www.wetourrobotics.com and on the Company's LinkedIn and X channels under Wetour Robotics and @WETO_IR_TEAM. Built on open research, trained on WETO's own architecture. Decoding continuous sEMG signals into hand pose builds on a fast-moving research frontier pioneered by Meta and others, and WETO trains directly on Meta's openly released emg2pose dataset. The Company is deliberately clear that the underlying capability is shared, open research rather than a proprietary first — its work is about what comes next: making that capability practical, affordable, and private enough to wear every day. Engineered for affordable, on-device deployment. Meta's research setup captures 16 channels at 2 kHz; Conductor targets a consumer-grade 8-channel band at 250 Hz. WETO bridges this gap in two stages: first, the model is pre-trained on the emg2pose dataset downsampled to 8 channels at 250 Hz to match Conductor's hardware, learning the core sEMG-to-pose mapping from a large, high-quality corpus; then it is adapted through transfer learning on WETO's own data, collected directly from the 8-channel, 250 Hz consumer band, so the model is fine-tuned to the exact sensor it will run on in the field. The architecture is a streaming, state-space (Mamba) model chosen for linear-time, constant-memory inference — designed to run fully on-device at the edge. In the demonstration, the model is evaluated on gestures it had not previo...

View stock analysis, news, and events for Wetour Robotics Limited