Speech recognition based chess system for visually challanged

Abstract

To overcome the obstacle of visual chess simulators, a system is proposed in which visually impaired individuals can practice by using only voice commands. In addition, this system will be powered using machine learning algorithms with the help of a publicly available repository of over five million games. This will make the user feel like he/she is playing against another human and not a machine. The existing chess systems use a variety of algorithms in order to choose its move. Most of these use tree traversals and the most common one is the min-max algorithm with alpha-beta pruning. Min-max algorithm finds the best move, and alpha-beta pruning prevents it from going into branches of the game tree that cannot yield a better result than previously traversed branches. Since the tree generated in a chess game is very deep and have a lot nodes, these algorithms examine the depth only to a certain amount. Generally, these algorithms are accompanied by an opening repository which bolsters the algorithm's efficiency. Since these algorithms are too strong for a visually impaired individual, a new approach is suggested to choose the computer's move. A publicly available repository of over five million chess games played by humans will be used to train the machine, initially. When the user makes his/her move, it will be converted into text and given as input to both the Minimax and the k-NN algorithms. The moves given as output by both these algorithms are compared using an evaluation function. The move with a higher score is chosen. The game continues till a decisive result is obtained.