Research On Fatigue Decision Based On Multi-dimension Data Fusion In Human Computer Interaction

The intelligent development trend of the "human-machine-environment" working model requires "humans" to be in a more focused working state,otherwise it may cause safety accidents.However,the fatigue state of people has become the norm in the work environment because of the fast paced life.Therefore,researching the wearable decision-making program of the human physiological state has important scientific significance and application value for ensuring the personal safety of the operator and improving his work performance.In the paper the "people" who perform perception,decision-making,and operation tasks in the human-computer interaction system is classified as "operation".This paper carries out comprehensive and in-depth theoretical and experimental research that focuses on some core problems about the extraction and analysis of fatigue-related physiological,eye-movement characteristics of human,and multi-dimensional data fusion technology.The research content is as follows:1)Extraction and Analysis of physiological signal characteristicsIn the paper we extract the multi-dimensional frequency domain features of the Electroencephalography(EEG)and the multi-dimensional time-frequency domain features of the Electrocardiography(ECG),and use the Relief method to selection feature and determine the EEG and ECG features that are highly related to human fatigue.Then we use the public data set to analyze the classification effect of the combination of EEG fatigue features,ECG fatigue features using the Naive Bayes classifier,K nearest neighbor classifier and support vector machine,and make the best performance combination as the corresponding fatigue classification model.Among them,the EEG fatigue classification model is used as the labeling tool of the simulation operation experiment,and the ECG fatigue classification model is used as the sub-learner of multi-dimensional data fusion fatigue decision-making model.2)Extraction and Analysis of eye movements characteristicsIn the paper,we propose two pupil detection algorithms:a method based on traditional image processing algorithms that uses dual Haar-like feature detectors to roughly locate the pupil position.the pupil detection algorithm obtains the precise position of the pupil through edge filtering,morphological pixel model and ellipse fitting;the other algorithm divides the pupil tracking task into pupil detection task and motion tracking task.The former uses a convolutional neural network model to detect the position and radius of the pupil,while the latter uses a long short-term memory network to obtain the current frame based on the pupil information of the previous two frames.Subsequently,based on the simulated operation dataset,we explore the correlation between eye movement saccade speed and human body fatigue,and establish the eye movement fatigue classification model as a sub-learner of multi-dimensional data fusion fatigue decision-making model.3)Multi-dimensional data fusion fatigue decision modelWe use Bayesian network fusion algorithm to achieve multi-dimensional decision-making layer data fusion,combine with the preliminary judgment results of the ECG fatigue classification model and the eye movement fatigue classification model to obtain the final fatigue decision result using the causal reasoning.Experimental results demonstrate that the multi-dimensional data fusion method proposed in this paper can achieves a good recall rate,and the overall detection performance is higher than that of the single feature classification model.