This paper presents a user interface designed to enable computer cursor control through hand detection and gesture classification. A comprehensive hand dataset comprising 6720 image samples was collected, encompassing four distinct classes: fist, palm, pointing to the left, and pointing to the right. The images were captured from 15 individuals in various settings, including simple backgrounds with different perspectives and lighting conditions. A convolutional neural network (CNN) was trained on this dataset to accurately predict labels for each captured image and measure their similarity. The system incorporates defined commands for cursor movement, left-click, and right-click actions. Experimental results indicate that the proposed algorithm achieves a remarkable accuracy of 91.88% and demonstrates its potential applicability across diverse backgrounds.

翻译：本文提出了一种通过手部检测与手势分类实现计算机光标控制的用户界面。研究团队构建了包含6720个图像样本的综合手部数据集，涵盖四个不同的手势类别：握拳、手掌、指向左侧和指向右侧。这些图像采集自15位受试者在不同环境下的姿态，包括具有不同视角与光照条件的简易背景。基于该数据集训练了卷积神经网络（CNN），用于精确预测每帧捕获图像的标签并度量其相似度。系统定义了光标移动、左键单击和右键单击的指令动作。实验结果表明，所提算法实现了91.88%的显著准确率，并展现出在多样化背景下的潜在适用性。