机器学习技术在游戏中的应用研究——解决鼠标轨迹识别问题  

Application on the Research of Machine Learning Techniques in Games

作　　者：刘天白[1] 张舒白[1] LIU Tian-bai,ZHANG Shu-bai（The 28 Research Institute of China Electronics Technology Group Corporation,Nanjing 210007,China）

机构地区：[1]中国电子科技集团公司第二十八研究所,江苏南京210007

出　　处：《电脑知识与技术》2011年第5期3100-3102,共3页Computer Knowledge and Technology

摘　　要：通过赋予游戏中智能体识别鼠标轨迹的能力,提高了游戏的可操作性,增加了游戏的趣味性。鼠标轨迹识别属于多分类问题,该文提出的解决方案是采用BP神经网络和支持向量机（SVM）求解8种鼠标轨迹的识别问题,并且完成了实验对比,封闭测试中两种方法对鼠标轨迹的正确识别率分别为93.75%和98.75%。Through attaching the ability of recognizing mouse track to agent in games,the controlling of games is simplified and the interesting of games is improved.Mouse track recognition is a multi-classification problem,which is resolved by BP neural network and Support Vector Machine（SVM）,the number of mouse track patterns is 8,according to contrastive analysis of BP neural network and SVM,the precision is 93.75% and 98.75% respectively.

关 键 词：游戏 轨迹平滑 鼠标轨迹识别 BP神经网络 支持向量机 

分 类 号：TP18[自动化与计算机技术—控制科学与工程;自动化与计算机技术—控制理论与控制工程]