Objective Reading behavior which is an important cognitive process can be realized by detecting the eye movements.This paper presents a detection algorithm which can identify reading activities effectively through the feature extraction of horizontal EOG.The detection algorithm can detect different reading behaviors and attention levels automatically.Methods First,the EMG and blink interference are removed from the horizontal EOG by mathematical morphology.Next,the features of the smoothed horizontal EOG are extracted by difference arithmetic.At last,the reading activities can be detected by the classification result of the features which are compared with actual reading activities.Results The analysis results of reading state and non-reading state in different periods show that the combination method of mathematical morphology with threshold value method has a strong anti-interference ability to the noise of blink and EMG.In this paper,a detection rate of 75% is achieved.Conclusions The combination method of mathematical morphology with feature extraction is effective for identifying the reading behaviors.This method can be applied to the recognition of daily reading activity.