Abstarct: Using sign language instead of mechanical keys for controlling equipment in industrial environments can create more safety as well as simplicity for the operator. To this end, information from operator’s hand is acquired by a visual camera and proper features will be extracted by using machine vision algorithms. The hand’s pattern then will be recognized from the extracted features by utilizing machine learning algorithms. Using this method, the operator’s environment can be controlled by hand’s pattern which works as a virtual key. Replacing mechanical keys with a virtual key is a new development toward performance optimization, cost reduction and safety increasement. In this thesis a new approach for recognition of hand’s pattern is proposed. To this purpose, after the acquisition of operator’s image, the hand’s part will be detected from the skin color. Due to the sensitivity of detection of hand’s region to the environment’s light we have taken advantage of fuzzy algorithms for sensitivity reduction. By extracting hand region, the contours will be obtained. We have used geometric equation of key points on the hand for extraction of hand’s pattern. After the extraction of points between fingers, a hypothetical circle will be obtained that move with the hand movement meaning that the operator can perform key patterns on the areas of the stage that has fewer interfering factors on the contour. Due to the requirements of KNN classifier and its little need to training data we use it for recognition of hand’s pattern. Empirical results from applying the proposed method on databaxse images show that it obtains great performance. The real time sensitivity rate or true positive rate of the classifier for 29 hand gestures, baxsed on 450 acquired image from the camera for each gesture, with an average of 96.40 for shiny sections and 94.33 for dim sections proves the accuracy of this method. The use of OpenCV C++ makes this algorithm work in real time.