1department of computer engineering, shabestar branch,islamic azad university,shabestar,iran
2University of Tabriz. Faculty of Electrical and Computer Engineering.
Analysis the video images is a research aspect that includes the video image research to provide the calculation tools which can help to identify physiological and pathological motions in the medical applications. The aim of the present study is to plan a system that automatically identifies the upper limb and its motions by analyzing the video images which distinguish the abnormal motions from normal motions. For this purpose, a database of hand gesture of 20 persons. In the obtained video images, blue, green, yellow, and red markers were placed on the 4 areas of tip of the fingers, wrists, elbows and shoulders, respectively. By image processing, the area related to each color was identified and the gravity center of the area was selected as the marker point. Finally, tremor of each marker was calculated individually. Also, based on the 0 to 9 grading, the intensity of the disease based on the tremor, was specified in each marker. To identify the tremor, we used the support vector machine (SVM) as the training model to categorize the videos. Based on the results obtained from SVM, the percentage of the accuracy related to the green marker was the highest accuracy (98.75%). The accuracy of the blue marker was 97.5% and accuracy of the yellow marker was 91.3%. Moreover, the least accuracy was related to the shoulder joint. . This marker separates the healthy persons from patients by accuracy of 86.8 % .