Olfactory Model Analyisis Based On Entropy

Olfactory bionic model studies plays an important role to understand theolfactory mechanism, and even the memory and learning mechanisms in the brain.KIII model is a representative of the bionic model for olfactory neural system, andprovides an effective object to study the olfactory system. But the complexity of theolfactory nervous system makes it difficult to analyze, so it is important to selectappropriate index to the study of olfactory model. In the thesis, KIII model is used asthe object of research and the concept of entropy in information theory is proposed asmethod for analysis of the KIII model, which provides a new perspective on the studyof the nervous system.In this thesis, the main work includes the following three aspects:First, the relevant theory is summarized. Using the position and role of varioustypes of cells in the olfactory system as the main line, three-tier structure includingthe olfactory epithelium layer, olfactory bulb layer and entorhinal cortex layer of theolfactory system is introduced. EEG is summarized and the common methods toanalyze EEG, including time-domain method, frequency domain, wavelet theory, areintroduced. A typical EEG, epilepsy EEG, is analyzed in detail.Second, the entropy is used to analyze KIII model. The entropy is the probabilityof specific information to be shown and can indicate the complexity of the signal. Themore complicated the signal, the larger the entropy.0and1are used as stimulus andinputted to KIII model. Different types of nodes in KIII model are chosen asanalytical objects using fuzzy entropy and approximate entropy. Experimental resultsbased on fuzzy entropy show that fuzzy entropy can reflect the difference of input andbe suitable to recognize input pattern. Experimental results based on approximateentropy show that the single stimulus only change approximate entropy of olfactorymodel output signal in a short duration, so the approximate entropy can be used toanalyze the adaptability of olfactory bionic model.Finally, by introducing the fuzzy entropy and KIII, the paper gives an explorationfor the identification of pattern recognition system using fuzzy entropy featureextraction. At the stage of feature extraction, according to the characteristics ofepilepsy EEG, each EEG signal is divided into segments of equal length, and then thefuzzy entropy for every segment is calculated respectively. Finally, all of the features are combinated together as the EEG signal’s feature vector. At the stage ofclassification, KIII model is used to classify EEG signals, at the beginning, normalEEG and epileptic seizure EEG are input to be the training set for the KIII model, thenthe KIII model is used to classify EEG. The experimental results show that the fuzzyentropy can characterize the EEG well. Especially, the KIII model has good ability toclassify epilepsy EEG based on the fuzzy entropy.