Research On Material Odor Information Processing And OCS Based On Bionic Olfaction

The bionic olfactory system also called Electronic Nose(E-Nose)is a type of odor information detection technology that was originated in the 1990s.It can be used to detect and identify material odors.In recent years,with the rapid development of electronics,computers and information technology,the application of bionic olfactory technology in defense,aerospace,biomedical,public security,food safely,chemical detection,and multimedia applications has attracted widely attention from researchers around the world.At the same time,with the in-depth research and rapid development of audio and video information,combined with network communication technology,we are witnessing tremendous expansion in auditory and visual functions in time and space.However,our research question in relation to olfactory system is;could human sense of smell broaden its time-space functionality through the study of Odor Characteristic Spectrum(OCS)?This paper fully analyzes the challenges faced by the bionic olfactory system in the detection,processing and identification of the material odor information based on the existing research foundations and achievements.It also conducts in-depth research on the detection and processing of material odor information in the bionic olfactory system.For the first time,the concept of OCS was proposed to investigate the general characteristics of the substance odor.It also carried out preliminary theoretical and experimental framework to explore answer to the research question and subsequent transmission and reproduction of odor information.The relevant work and main results of this article are summarized as follows:(1)For the bionic olfactory system,the baseline change(electronic drift)of the metal oxide gas sensors due to temperature sensitivity causes the acquisition of detected data to be distorted(different result at repeated reading).A signal pre-processing method based on Z-distribution normalization was proposed,to effectively improve the data drift The experimental results show that the proposed method effectively improved the raw data by eliminating noises generated due to the sensor drift at room temperature,and provided a stable and effective data source for subsequent signal processing and feature extraction.(2)The feature extraction of material odor information is the basic problem of signal processing in bionic olfactory system.This paper proposes two kinds of mathematical characteristics based on curve simulation and statistics to solving the data map collected by PEN3 electronic nose system,which can effectively characterize the intrinsic characteristics of the measurement map.Effectively reduce the data processing dimension.The test results show that we proposed an odor map based on the two mathematical characteristics mentioned.It is more intuitive to distribute in the feature space than the original data.(3)The feature reconstruction of material odor information is the key issue in signal processing of bionic olfactory systems.Limited by the effects of metal oxide sensor materials and manufacturing processes(cross-sensitivity).The data map are high-dimensional in the data space and has overlapping effects,based on the PEN3 electronic nose system.In this paper,a linear discriminant method based on Kernel Discriminate analysis(KDA)was proposed to deal with dimensionality reduction and feature extraction of high-dimensional data maps.The simulation results show that the proposed method can effectively extract and classify the feature data of 4 types of industrial gases.(4)In order to further explore the reconstruction of high-dimensional data map features collected by the bionic olfactory system,this paper proposes the Selective Local Linear Embedding(SLLE)algorithm based on feature selection.It was used to reduce dimension and classify high-dimensional data collected by the PEN3 electronic nose system.The dimensionality reduction features are mapped to two-dimensional and three-dimensional feature spaces.It can realize the effective detection and classification of odor information,combined with the Euclidean distance classifier and solve the problem of dimensionality reduction and identification of high-dimensional complex data in bionic olfactory systems.The simulation results show that the proposed method can classify and identify the Aucklandia Lappa materials of eight different origins.Compared with the classic linear dimension reduction and feature extraction methods,the feature space distribution and recognition rate are significantly improved.(5)The selection and characterization of material odor characteristics are the frontier issues in current bionic olfactory signal processing research.Due to the lack of a unified feature characterization model,the problem of odor information features is not universal.This article proposes the concept of OCS for the first time,and explains the physical definition,which specifically includes the characteristic information such as the odor name,composition information and concentration ratio.The goal is to describe the general characteristics of the odor of a substance.The method of extracting the odor of material was discussed in detail,and a superposition mapping analysis(SMA)algorithm was proposed to achieve the OCS static parameters.At the same time,a concentration estimation model was proposed to estimating the OCS dynamic parameters.It provides an exploration for the subsequent theoretical improvement and application research of OCS.