A Smart Sensor-based Gesture Recognition System for Media Player Control

With the rapid advancement of technology in the current era, mobile and wearable devices have attracted the attention of end users. Smartwatches available in the market have embedded movement sensors whose potential applications have not been explored to a great extent. This study proposes a gesture recognition module using a smart sensor Metawear which can potentially be used to control various functions of a media player. Inertial sensors data from 9 subjects for 12 different gestures (3 trials for each gesture) are acquired. Fifteen different features are extracted from the acquired data. Wrapper method for feature selection is applied on the extracted features which select features from six features groups among the fifteen extracted feature groups. Classification of gestures is performed using three different classifiers i.e., random forest (RF), Naive Bayes (NB) and k-nearest neighbors (kNN) algorithm. The best classification accuracy of 85.80% is achieved by using a fusion of features extracted from accelerometer and gyroscope data by kNN classifier. Experimental results show that the proposed scheme outperforms the existing schemes in literature in terms of classification accuracy for the larger number of gestures.