本文建立了一个基于BP人工神经网络的手部气味识别模型，并对网络参数进行了选择优化。人体手部气味经样品采集、浓缩后，用气相色谱–质谱联用仪对其进行分析获得手部气味轮廓图，并利用逐步判别分析法提取了手部气味轮廓图的47个特征变量(有机化合物)。在此基础上，构建了一个47 × 10 × 1的BP网络，其隐含层和输出层的传递函数分别为tansig和logsig，训练函数为trainrp。该网络能正确区分不同性别的手部气味。
A Recognition Model of Hand Odor Based on BP Artificial Network
Abstract: A recognition model has been developed based on Back Propagation Artificial Network and its net parame- ters were optimized. Hand odors were sampled, concentrated and then analyzed with Chromatography-Mass Spectrometer, resulting in hand odor profiles whose feature variables (47 organic compounds) were extracted and reduced by means of stepwise discriminant analysis. In sequence, a BP network, in which the structure was 47 × 10 × 1, the transfer functions for hidden layer and output layer tansig and logsig, respectively, and the training function trainrp, was pro- posed. The experiment demonstrated that the network was able to specify whether an odor sample had been from man hand or woman hand.
文章引用: 龙成生 , 王 辛 , 吴德华 , 张汇东 , 宋珍华 (2012) 基于BP网络的手部气味识别。 计算机科学与应用， 2， 57-60. doi: 10.12677/CSA.2012.22011
Copyright © 2020 Hans Publishers Inc. All rights reserved.