A noise-tolerant human–machine interface based on deep learning-enhanced wearable sensors
摘要
Wearable human–machine interfaces with inertial measurement units (IMUs) are widely applied across healthcare, robotics and interactive technologies. However, extracting reliable signals remains challenging owing to motion artefacts in real-world environments. Here we present a human–machine interface capable of tolerating diverse motion artefacts through deep learning-enhanced wearable sensors. The system integrates a six-channel IMU, an electromyography module, a Bluetooth microcontroller unit and a stretchable battery, enabling wireless capture and transmission of gesture signals. A convolutional neural network trained on a composite dataset of gestures and motion artefacts extracts robust signals, while parameter-based transfer learning improves the generalizability of the network across users. A sliding-window approach converts the extracted gesture signals into real-time, continuous control of a robotic arm during dynamic activities such as running, high-frequency vibration, posture changes, oceanic wave motion and combinations of these. This work demonstrates the potential of wearable human–machine interfaces for complex real-world applications.