Combines the Raspberry Pi with YOLO v3 target detection algorithm

and a new hexapod robot humancomputer interaction mode is proposed

which enables the hexapod robot to complete five different actions such as forward

backward

left turn

right turn

gait switch. Among them

indoor and outdoor gesture recognition can reach95. 08 % and 89 %

respectively on the test set. Compared with traditional algorithms based on artificial extraction features and other CNN-based algorithms

it has higher recognition accuracy and stronger robustness.