Human computer interface using hand gesture recognition based on neural network

Gesture is one of the most vivid and dramatic way of communications between human and computer. Hence, there has been a growing interest to create easy-to-use interfaces by directly utilizing the natural communication and management skills of humans. This paper presents a hand gesture interface for controlling media player using neural network. The proposed algorithm recognizes a set of four specific hand gestures, namely: Play, Stop, Forward, and Reverse. Our algorithm is based on four phases, Image acquisition, Hand segmentation, Features extraction, and Classification. A frame from the webcam camera is captured, and then skin detection is used to segment skin regions from background pixels. A new image is created containing hand boundary. Hand shape features extraction, are used to describe the hand gesture. An artificial neural network has been utilized as a gesture classifier, as well. 120 gesture images have been used for training. The obtained average classification rate is 95%. The proposed algorithm develops an alternative input device to control the media player, and also offers different gesture commands and can be useful in real-time applications. Comparisons with other hand gesture recognition systems have revealed that our system shows better performance in terms accuracy.