Gait Feature Perception and Identity Recognition Under Visual-Tactile Dual Modality

Gait recognition technology, as a key branch in the field of identity recognition, has significant value but also faces challenges such as discomfort from wearable devices, the need for synchronized processing of gait feature time series data, and susceptibility to environmental influences in image-based methods. A fusion method combining tactile and visual information is proposed to improve recognition effects when sensors or visual information are limited. Not only is a fusion gait dataset constructed, but identity recognition experiments are also conducted on flexible strain sensor data through Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) CNN-LSTM networks. The effectiveness of the visual-tactile fusion method in enhancing recognition accuracy is verified through decision-level fusion strategies.