Gesture recognition based on improved SVM-RW-KNN

In the current field of artificial intelligence, most gesture recognition research are based on original KNN and SVM. Alternatively, gesture recognition could be on a single KNN or SVM algorithm. Therefore, the gesture recognition accuracy is not satisfactory. In this paper, a K-nearest neighbor (KNN) classifier based on weight allocation is proposed, with the finger feature importance of random forest as the weight (RW-KNN) to improve the accuracy of gesture recognition. Meanwhile, support vector machine (SVM) classifier has been parameter optimized and combined with RW-KNN (SVM-RW-KNN) for gesture binary classification on gesture dataset which is collected by Leap Motion, can further improve the accuracy of gesture recognition.