摘要
提出将支持向量机应用到气体种类识别的研究中,并建立小波分解提取特征量和支持向量机识别气体种类的气体定性分析模型。通过小波分解提取半导体气体传感器在温度调制下的动态响应特性的特征量,分别使用不同核函数和不同结构的支持向量机建立判断特征量与气体种类的模型。实验结果说明使用支持向量机进行气体成分定性识别的效果优于同结构的神经网络,且对支持向量机自身结构的选择不敏感,适合于对多组分气体定性分析研究。建立的模型在分辨力为13ppm(对CO)和15ppm(对H2)的条件下,对单一氢气、一氧化碳及其混合气体的识别率可达98%,适合于工程应用。
A new classification model based on support vector machine and wavelet decomposition for gas identification is proposed. In this model, the wavelet decomposition is applied for feature extraction from a single temperature modulated semiconductor gas sensor, and the support vector machine is introduced for pattern recognition between the feature variables and gas patterns. Experiments of support vector machines with different kernel functions or structures were used to test the performance. Practical results show that with feature variables extracted from wavelet decomposition, the support vector machine model possesses better recognition ability than the traditional neural network does. The classification ability of support vector machine can be 98% for single CO, H2 and their binary mixture conditions at a precision of 13ppm （for CO）and 15ppm （for H2）, and it is suitable for industrial application.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2006年第6期573-578,共6页
Chinese Journal of Scientific Instrument
关键词
气体识别
支持向量机
小波分解
温度调制
气体传感器
模式识别
Gas identification Support vector machine Wavelet decomposition Temperature modulation Gas sensor Pattern recognition
作者简介
葛海峰 1981年生 男 硕士研究生 主要研究方向为信号处理 故障诊断 无线传感器网络 E-mail：xjtu-pigeon@yahoo．com．cn
