Classication of Student Activities Based on a Sequence of Images from a Single Camera

摘要

In an active learning environment, activities of a student is crucial to his or her learning achievement. However,keeping track of the student activities by teaching staff alone is almost impossible. Hence, using technology forsuch tedious but important job has become attractive or even necessary. Focusing on such environment, thispaper proposes a method of classifying whether a student is writing, or reading, or working on a task such asdoing an experiment, based on sequential image frames from a single camera. For each frame, an area includingthe student is cropped out using a background subtraction and thresholding. Then, using the skin detectiontechnique, face and hands of the target student are detected. Such face and hand areas in a sequence of frames arecombined as a Gait Energy Image (GEI), which is being used as feature images for the classification, in whichthe Principal Component Analysis is applied. A score of the PCA, in which each row serves as an observedsample, is taken as a feature. In addition, another score of the PCA, in which each column serves as an observedsample, is taken to be another feature. Using the support vector machine, the two features are used rst toclassify whether a student is "reading" or "not reading." Then, a sample of "not reading" is further classiedinto either "writing" or "doing experiment." Based on a sequence of simulated activities, the proposed methodcan classify between "reading" and "not reading" with 93% accuracy, while the classifying between "writing"and "doing experiment" achieves 90% accuracy.