Performance of Classifier for Gesture Recognition Using Machine Learning Techniques

Verbal communication is the most prominent way of exchange of information between people. In the case of deaf and dumb people due to verbal and hearing impairment, the communication form followed is sign language, where different gestures are followed. The normal people require conversion, and due to this, communication barrier occurs between normal and deaf-dumb people. This leads to the development of a system with a translation device that could enable mute people to convey the message efficiently. The proposed device consists of flex sensors and MPU6050 (combination of accelerometer and gyroscope) unit to record hand movement. The recorded values of the hand movement is processed using machine learning algorithms to classify and predict the gestures accurately. In this work, an attempt has been made to implement different ML techniques such as random forest, decision tree, K-nearest neighbour, and support vector machine. It is found that the classification accuracy using random forest is 98.5%, which is better when compared to other ML techniques considered. The proposed system is better suitable for communication with normal people. .