Research On Feature Extraction And Recognition Algorithm Of EOG Signal

Nowadays, the processing of bioelectric signal has becoming a hotspot in the field of computer application. Based on the bioelectric signal, the Human-Computer Interaction (HCI) is an interactive system which establishes connections between human body and the computer or some other electrical instruments, to accomplish preconcerted actions through extracting bioelectric signal of human body and further transferring them into corresponding peripheral equipments orders. A revolution in the field of traditional interactive modes has been triggered by the studies on such an interactive system based on bioelectric signal.EOG signal is weak bioelectric signal which is produced in the process of eye movements. And those signals can be detected by electrodes pasted around eyes. With the advantages of high amplitude, being convenient for detection, being relatively easy for processing and so on, the system of HCI based on EOG signal will undoubtedly possess wide developing prospects.This thesis mainly studies the feature extraction and recognition algorithm of EOG signal to get a series of high effective and transplantable feature parameters. Based on this principle, some tasks have been accomplished:1. Collection and pre-processing of experimental data:In the situation of laboratory, the experiment of collecting EOG signal of several experimenters has been designed and a large number of initial EOG data have been acquired.2. Feature extraction of EOG signal:an EOG feature extraction algorithm based on Linear Predictive Coding (LPC) technology has been proposed in the thesis. Namely, some LPC coefficients, which extracted from the pre-processed EOG signal (pre-processing procedures include:filtering, framing, computing short-time energy, endpoint detecting and so on), are used as feature parameter of EOG signal. To get dynamic information of EOG signal, the first order difference LPC (ΔLP)coefficients and peak value information are further extracted. As a result, a combination which includes LPC coefficients,ΔLPC coefficients and peak value has been used as feature parameters of EOG signal.3. Pattern recognition algorithm of EOG signal:the multi-layer feed-forward network of BP algorithm has been studied on EOG signal classification, and some key parameters have been acquired by many different experiments. Under the environment of MATLAB 7.0, the simulation experiment has been designed and realized; furthermore, the algorithm proposed by this thesis has been validated by identifying the EOG signal.4. Communication with external equipment:The pattern recognition of EOG signal is made through VC++6.0 platform. By adopting the serial communication technology, an on-line control system has been realized based on EOG signal.Through making a study on the collection and pre-processing, feature extraction and pattern recognition as well as software realization of EOG signal, and by designing and accomplishing an uncomplicated HCI system with the basis of EOG, this thesis has laid a foundation for more effective HCI system. In the part of conclusion, the author of this thesis makes a summary of this study and also a prospect for the work in future.