摘要
提出了一种利用BP神经网络仿真、利用贝叶斯决策修正仿真结果的人脸检测方法。讨论了单纯使用BP神经网络作人脸的检测判定的不足,并在此基础上提出利用贝叶斯决策对神经网络的仿真结果进行第二次判定的方法。共使用MITEx人脸库的4 000个人脸与非人脸图像进行实验分析,正确率平均提升了3.63%,表明了神经网络的良好判定性能和使用贝叶斯决策进行修正的有效性和必要性。
Putting forward a face detection method which make use of BP neural network simulation and fixed the results of the simulation by Bayesian decision. Discussed the deficiency in face detection of BP neural network which was single used, and put forward a re-decision method by Bayesian decision. There were 4 000 face and non-face images which came from MITEx face database were used to complete the experiment, and the correct rate raised 3.63%. The results show the well distinguish effect of neural network, as well as the efficiency and necessary of Bayesian decision.
出处
《计算机应用研究》
CSCD
北大核心
2007年第8期198-200,共3页
Application Research of Computers
基金
国家自然科学基金资助项目(60372072)
关键词
人脸检测
反向传播神经网络
仿真
贝叶斯决策
face detection
BP（back propagation） neural network
simulation
Bayesian decision
