Recognition of Turkish Command to Play Chess Game Using CNN

Abstract

A platform has been created that allows playing chess with Turkish voice commands. The aim of this study is to enable individuals with limited movement abilities as a result of congenital reasons or a certain disease or accident to play chess and perform a social activity without the help of another person, and to be rehabilitated at the same time. It consists of three parts: Chess module, Human-computer interaction module and Artificial Intelligence module. 29 words have been determined to provide movement in the game on the platform. Voice recordings from 151 people, 86 men and 65 women, were used. Feature selection was made on 43790 voice recordings by using mel frequency cepstral coefficients (MFCC) and gammatone cepstral coefficients (GTCC) methods. The results obtained were classified using the traditional CNN model. The data obtained after using MFCC and GTCC methods were used as inputs in the CNN model. In addition, the data obtained by the two methods were combined and trained in the model. Depending on the methods used in the created model, 83% to 85.9% results were obtained. It was determined that the results obtained using the MFCC method were more successful.