Most of the existing gesture recognitionapplications are based on specific devices

such as Kinect

Leap Motion

etc.Because of its price and practicability

it can not be widely used. Monocularcamera is the most common accessory of computer or mobile devices. With thecorresponding gesture recognition and interaction algorithm

gestureinteraction can be carried out. In this paper

the algorithm of gesturerecognition and interaction based on monocular camera is studied. The monocularcamera captures the image

uses the skin color to segment the image

andremoves the noise points according to the contour area to get the binary imageof the gesture image. Then

the matrix of binary graph is introduced into theneural network as a parameter for deep learning pattern recognition. Throughconvolution and fullconnected network structure

as well as optimization ofparameter selection and prevention of overfitting

combined with fingertipdetection and elimination algorithm

the recognition accuracy is improved. Inthe aspect of gesture interaction

the motion on the plane of gesture image isjudged by the change of gesture center of gravity

and the displacement isdetected to prevent the sudden change of gesture image position caused by theinterference. In the front and back movement of gestures

the area of differentgestures is compensated for and adjusted by the size of gesture images

so asto reduce the interference caused by the area changes caused by differentgestures.