信号的特征提取和模式识别方法,在实现准确的电子鼻气体定性分析中尤为关键,本文提出了基于AR信号处理和KII模型的嗅觉识别算法.将传感器信号分为:上升期和稳定期两部分,对上升期信号提取斜率作为特征；对稳定期信号,进行AR建模来提取特征.在电子鼻的模式识别算法上,利用KⅡ模型对气味信号进行分类.该方法充分利用了AR信号处理在信号表示方面的有效性及降维优势、KⅡ模型在模式识别方面的优越性.仿真将该方法与BP网络、AR_BP算法及单KⅡ网络进行了比较,结果表明,AR信号处理技术可以很好的提取特征,并与KⅡ建立相关的数学模型,将AR信号处理技术应用到电子鼻系统中是可行的,且具有更高的识别率.%Signal characteristics extraction and pattern recognition, in accurate gas analysis is critical for electronic nose, Olfactory recognition algorithm based on AR signal processing and KII model ( AR_KII) is proposed. Sensor signals are divided into two parts: rising and stable period, extract the rising periods signal slope as features; stable signal AR model to extract features. On pattern recognition algorithm of electronic nose, to classify the odor signals by KII model. This method made use of AR signal processing to signal that the effectiveness of and dimension reduction advantages and superiority of KII model in terms of pattern recognition. Compare AR_BP with the BP network, AR_BP, KII and KII_BP through Simulation, experimental results show that the AR feature extraction of the signal processing techniques work well, and build related mathematical models and KII, AR signal processing techniques applied to the electronic nose system is feasible and has a high recognition rate.
展开▼
