Gesture Recognition System based on YOLOV8 Convolutional Neural Network

Gesture recognition, a pivotal area within human-computer interaction, has garnered significant interest in recent times. However, it takes time and effort to identify the various human gesture images in reality and consumes a lot of manpower and material resources. In order to solve the above problems, this paper applies the YOLOv8 algorithm, an object detection algorithm based on convolutional neural network, to the research of gesture recognition system. Firstly, the initial data sample set is constructed by using the public open source dataset Ha-GRID, due to the original dataset is too large, this paper adopts eight of the common gesture samples by random sampling method. Where the initial dataset contains a total of 63945 human gesture images. Secondly, by studying the differences in the features of various gesture images, this paper adopts the literature research and analysis method to select the latest YOLOv8 model, which has a faster detection speed and can better realize the efficient and accurate recognition and detection of the common human gesture images because it adopts a more efficient network structure and smaller model parameters. And after this experimental results show that the final recognition accuracy of YOLOv8 model is up to 99.4%, the recall rate is 99.1%, and the average accuracy is also up to 99.5%.