Nianxiang Tech, founded by a brain computer interface veteran with stints at a Chinese brain implant firm and the Chinese humanoid robot startup Agibot, raised about 10 million yuan to build a surface electromyography (sEMG) wristband and seed a Chinese muscle signal dataset for…
Picture a commuter on a noisy subway, flicking their wrist to scroll a short video on a pair of smart glasses, with no touch screen, no keyboard, no voice command. The trick is a band on the wrist that reads the faint electrical signals the muscles fire when the hand moves, translates them into finger and wrist gestures, and hands those gestures off to the phone, PC, or smart home device over Bluetooth. That is what 念象科技 (Nianxiang Tech) is building with its Omniband prototype, and the Beijing- and Shanghai-based startup just closed an angel round of nearly 10 million yuan (about $1.4 million at recent exchange rates) to push the device from the lab toward a real product.
The bet hinges on a quiet technical threshold. In July, researchers from Meta's Reality Labs published a paper in Nature showing that surface electromyography, the technique of measuring muscle electrical signals on the skin rather than reading the brain directly, can recognize hand gestures across different users without per-person calibration, provided the model has been trained on signals from roughly 100 people. That number matters because individual variation in muscle anatomy and electrode placement had been the bottleneck for a decade. Meta's own blog post framed the work as a step toward gesture control for its Orion augmented-reality glasses, and The Register called it a UI breakthrough for neural wristbands. 念象科技's founder Dr. Wang Yi (王译) read it as a market opening, not a Western research curiosity.
Wang is not the typical hardware CEO. He holds a PhD in brain-computer interfaces from the University of Auckland, currently serves as vice chair of the National Brain-Computer Interface Industry Alliance (国家脑机接口产业联盟), and was named a Shanghai Magnolia (白玉兰) talent, according to a 36Kr exclusive that broke the funding news this week. His résumé spans the three technical paths the field has argued about for years: invasive brain implants (as former chief scientist at Yingmai Medical / 应脉医疗), scalp EEG headsets, and the wrist-based sEMG approach that Omniband now uses. He was also previously R&D director at Agibot (智元机器人), a Chinese humanoid-robot startup. That breadth is what makes the consumer-wristband starting point credible rather than a leap of faith, because Wang has personally seen why the medical and EEG routes stalled.
Omniband itself is still an engineering prototype. The wristband continuously estimates the dynamic angles of all 20 finger and thumb joints from wrist neuromuscular signals, supports airborne handwriting, and exposes a Bluetooth HID interface so phones, PCs, smart glasses, and smart-home hubs can treat it like an external controller. Gaming and short-video control are the most polished use cases so far; the full out-of-the-box experience, where you put it on and it just works, has not been achieved. To handle real-world noise from arm motion and sweat, the team uses differential electrodes and a hardware structure designed for stable skin contact, runs custom sEMG filtering and signal-separation algorithms, and cross-validates multiple sensor streams inside the model.
The hardware is half the story. The longer-horizon asset is data. Wang wants to build a large-scale Chinese sEMG dataset with labels for hand pose, muscle force, and object interaction, and treat it the way ImageNet treated image recognition, as a public foundation that anyone training embodied AI, the kind of AI that learns from physical-world interaction rather than text, can build on. Today's Chinese embodied-AI efforts rely on foreign datasets built mostly from Western hands and Western motion-capture rigs; a domestic dataset with the same breadth would be both a research shortcut and a strategic moat.
The round, disclosed via the 36Kr interview, is led by Yongjun Xingmang (永珺星芒), with Pudong Venture Capital (浦东创投) and Yicun Capital (一村资本) co-investing. Capital will go to product R&D, team hiring, and the dataset buildout. The company was founded late in 2025.
The honest caveats sit alongside the promise. Meta is not standing still: it acquired CTRL-Labs in 2019 and its CTRL-Labs lineage inside Reality Labs remains the deepest industrial effort on this stack. The scaling-law claim that makes cross-user sEMG work, roughly 100 users of training data, comes from one paper in one lab; replication across populations, electrode designs, and skin types is still open. And Omniband's "ImageNet of Chinese sEMG" is a goal, not a dataset. What the angel round really buys is a credible Chinese team positioning itself at the moment when the underlying signal-processing problem has just become tractable, and when the consumer devices that would actually host a neural wristband, smart glasses in particular, are starting to ship in volume.