The invention belongs to the crossing field of flexible sensing and
machine learning, and provides a wearable thumbstall
system integrated with a self-monitoring learning
algorithm, which can quickly adapt to a new task and a new user under the condition that only a small amount of data is used. According to the thumbstall, the two telescopic flexible sensors are integrated at the two joints of the
thumb respectively, so that fine actions of the
thumb can be captured with high precision. The flexible sensor is constructed based on an ionized water
gel electrode of a periodically distributed layered
triangular prism array structure, and a
dielectric layer formed by a conductive silver
nanowire and an Ecoflex
composite material is sandwiched, so that excellent stretchability and electrical
signal response capability are realized. In the aspect of
algorithm design, a comparative learning model based on
time sequence information is provided, discriminative characterization can be automatically extracted from random
thumb motion signals, and therefore accurate recognition of multi-fingertip interaction tasks is achieved. Experimental results show that the
system achieves the recognition accuracy exceeding 92% in the two tasks of direction
gesture recognition and virtual key input. More importantly, the
system supports free switching among tasks, the model does not need to be trained again, the practicability and the
automation level of the system are remarkably improved, and the system has the potential of being popularized to wider man-
machine interaction scenes.