Research On Electronic Nose Based On Hierarchical Temporal Memory Model

The electronic nose is a kind of bionic instrument used for gas detection, which generally consists of an array of cross-sensitive gas sensor and an appropriate pattern recognition algorithm. Compared to other gas analysis method, like specialized sensory panel and modern chemical analysis instruments, electronic nose technology has the advantages of simply, objective and low-cost.The researchers have tried a variety scheme of the data acquisition methods and pattern recognition algorithms to achieve the electronic nose systems, among them the one that gas sensor array working with the artificial neural network method performs well. The key research of electronic nose focus on two aspects:cross-sensitivity gas sensors and pattern recognition algorithms. The array of gas sensor is the device of acquisition, and its response sensitivity and stability determines the performance of the entire system. Pattern recognition algorithm is the key factor deciding the ability of electronic nose system, many artificial intelligence methods, including artificial neural networks has been applied to the electronic nose system, but the improvement of the recognition rate and performance is still the research focus of electronic nose. In this paper we design an electronic nose system, and apply a new pattern recognition algorithm named HTM to artificial olfactory identification system, to detecting and discriminating gases.This paper uses sensor array and HTM algorithm to design electronic nose system. We mainly study a new artificial intelligence algorithms named Hierarchical Temporal Memory, and apply it to electronic nose as a bionic information processing technique. The main content of this paper are as follows:Firstly, the data acquisition device of the electronic nose experimental system is designed. This system includes the sampling of gas, the control of reaction chamber, and the drive of sensor array as well as data acquisition process. We also design the pre-processing and feature transformation scheme of the signals.We design the sampling of gas, the configuration of sensor array, data acquisition circuit design as well as a PC data acquisition software.Then, we study the HTM cortical learning algorithms in detail, containing HTM principle of hierarchy, encoding strategy, the capability learning and predictive. We implemented an artificial olfactory recognition system to achieve automatically detecting and discriminating gases based on HTM. Furthermore, the olfactory recognition system is applied to gas detection. The experimental results show that the algorithm based on the HTM has excellent performance in classification of gases. Its robust generalization capability is suitable for electronic nose applications to process the signals of electronic nose.