为了提高红外步态识别精度的目的,采用分别基于小波描述子特征的模糊分类器识别和基于体形平均灰度图特征的贝叶斯分类器识别,再进行基于遗传算法和BP模糊神经网络的多分类器融合识别的新方法。做了基于中科院红外步态数据库的识别仿真实验,获得识别率、等错误率和累积匹配分值的实验数据及对比结果,得到多分类器融合识别比单分类器识别提高约10%识别率,降低约10%等错误率,完全收敛阶数提高1倍多的结论。具有识别精度高、收敛速度快的特点。
A new algorithm is proposed in order to improve the precision of the infrared gait recognition.The new method adopted the fuzzy classifier recognized by characteristics of the wavelet descriptors and the Bayesian classifier based on shape features of average grayscale respectively,and then performed the fusion recognition of multiple classifier based on genetic algorithm and BP fuzzy neural network.The recognition simulation experiment was made based on the infrared gait database of the Chinese Academy of Sciences.The comparison results and experimental data about the recognition rate,the error rate and the cumulative match score were gained.The conclusion shows that the multiple classifier fusion recognition increased about 10% at the recognition rate,reduced about 10% at the equal error rate,increased 1 times more at the complete convergence order number than the single classifier recognition.The characteristics of high accuracy and quick convergence are obvious.