Study On Barrier-free Human-computer Interaction System Based On EOG Encoding

A barrier-free human-computer interaction technology uses biological signals instead of the traditional hardware interfaces to meet the interaction and communication of users and external devices(computers,etc),which is designed to help people with disabilities to communicate with the outside world and integrate into the modern information society.Most information from the outside transmits to the brain through the eye.Different eye movements of different exercising forms can produce corresponding EOGs,so EOGs can be used as control signals of human-computer interaction.Ocular device is simple and portable,and the waveform characteristics of EOGs are obvious.Such advantages make human-computer interaction system based on EOG greater significant and practical valuable.This paper studies on the barrier-free human-computer interaction based on EOG,proposes collection experimental paradigm of EOG,and analyzes the offline information about eye movements and blink of different subjects,and compare the simulation results of different pre-treatment and feature extraction and classification algorithms of EOG,and discusses the reasons for differences between subjects' EOGs,and proposes a signal processing mechanism,which can regulate the system parameters to adapted individual differences.Finally,this paper compares the results under different algorithms and designs human-computer interaction with these algorithms through experimental results.Online experiments test the robustness,reliability and accuracy of this system.The following work is carried out in this paper.Firstly,the study on the eye movement pattern is expanded to eight directions(up,down,left,right,left-top,leftbottom,right-top,right-bottom)to increase the number of control signals of the system.As for the preprocessing of EOG,this paper uses values of each segment of eye-resting in front of each eye movement to compensate the baseline of EOG,and compares conventional Butterworth low-pass filter with wavelet denoising.A new adaptive median filter is proposed to remove unconscious blink interference.End-point detection is proposed by the three thresholds based on short-term energy normalized.On the part of feature extraction and classification of EOG,a feature extraction method combined least squares polynomial fitting parameters with linear prediction coefficient and a method based on multi-resolution analysis is proposed.Then this paper compares the results of using minimum output coding support vector machine to identify the direction of eye movements with three feature extraction methods above,and programs the algorithms mentioned above by MATLAB and LABVIEW to confirm them validity.Finally,a typing system based on eight eye movement directions is designed with the combination of different algorithms.This paper uses the online data transmission rate and accuracy of results to assess the performance of this system.The ocular recognition rate can reach 88%,and online character output rate could reach 13.75 characters/min,which is superior to that of ocular type system.And then adds entertainment and monitoring function to this system by using reserved codes.