International Journal of Computer Applications (0975 – 8887) Volume 83 – No4, December 2013 5 Hand Motion Tracking for Alphabet Recognition using ANN Shivganga Udhan ME student, Computer Department, Sinhgad COE, University of Pune P. R. Futane Professor & Head, Computer Dept, Sinhgad COE, University of Pune R. V. Dharaskar Professor & Director, Matoshree Pratishthan, S.R.T.M. University, Nanded ABSTRACT With the Objective of making communication process easy between handicap people and computer this paper describes a very basic method for recognizing the alphabets from hand motion trajectory. This method uses Artificial Neural Network (ANN) for alphabet recognition from gesture path (motion trajectory). This is done in three main stages; preprocessing on video input, feature extraction and classification. In first stage, preprocessing, sticker of particular colour placed on users hand is detected using color information. After detection of required colour to be traced, its motion trajectory which is also called as gesture path will be determined by tracking the sticker placed on hand. In the second stage, features are extracted from hand motion track which can be further used for training & testing ANN. In the final stage alphabet is recognized by using extracted features of gesture path. General Terms Motion Tracking, Human Computer Interaction, Pattern Recognition, Hand Written Character Recognition, Security and Algorithms. Keywords Hand Motion Tracking (HMT), Artificial Neural Network (ANN) and Character Recognition. 1. INTRODUCTION Communication is a major part of human life and is a social activity. Basically, communication is a transfer of information. Based on the channels used for communicating, the communication process can be broadly classified as verbal communication and non-verbal communication. Communication without words is called as Nonverbal communication. And the verbal communication includes written and oral communication. Sending written message is formal way of communication and is original form of the message. Verbal communication is very advantageous in the situation where speech commands are disturbing, environment is noisy, quantitative information is to be communicated etc. Human- computer interaction is also one of the emerging field which uses form of verbal communication to make the human interact with computer as they interact with each other. Also there can be physiological impairments like physical disorders, which make writing impossible. But written communication is must in many cases. To bridge up the communication gap between normal people and physically handicapped people, and also between physically handicapped people and computer, there is requirement of some system which make it possible to have formal written communication by physically handicapped person with other people or computer. Nowadays many systems are working with vision based approach for human motion tracking, but very few work is been done to have formal communication without any physical input device. So, there is the need of a system that can accept the video of human hand or any body part as an input trace out the path & process it to identify the alphabet from that motion track. Further, it should display a corresponding capital alphabet. 2. RELATED WORK An overview of feature extraction methods for off-line recognition of segmented (isolated) characters is represented by Oivind Due Trier et.al [5]. Vinita Datt and Sunil Datt have proposed a system capable of recognizing handwritten characters or symbols; they build a system that recognizes handwritten characters. The system should be such that it should be able to handle transformation of translation, scaling or a combination of both. The objective of this is to bring out accurate results even for images with noise in them. [9] M Elmezain, et.al, has presented a method to recognize the alphabets from a single hand motion using Hidden Markov Models (HMM) [3]. A simplified approach to recognition of optical or visual characters is portrayed and discussed by Shashank Araokar [6]. Nisha Vasudeva et.al, have introduced a method in which each image character is comprised of 30×20 pixels. Features extracted from characters are directions of pixels with respect to their neighbouring pixels. These inputs are given to a back propagation neural network with hidden layer and output layer. They have used the Back propagation Neural Network for efficient recognition where the errors were corrected through back propagation and rectified neuron values were transmitted by feed-forward method in the neural network of multiple layers. [4] Kauleshwar Prasad, et.al, focused on recognition of English alphabet in a given scanned text document with the help of Neural Networks. Using Matlab Neural Network toolbox, also they have attempted to recognize handwritten characters by projecting them on different sized grids. They have used character extraction and edge detection algorithm for training the neural network to classify and recognize the handwritten characters. [2] Jianbo Shi and Carlo Tomasi have proposed a feature selection criterion that chooses the correct features which will give optimal use of those features. It is based on how the tracker works, and features that do not correspond to points in the world are also identified. The algorithm used is extension to the previous Newton-Raphson style search methods to work under affine image transformations [1].