K-mer-Based Human Gesture Recognition (KHGR) Using Curved Piezoelectric Sensor

Recently, human activity recognition (HAR) techniques have made remarkable developments in the field of machine learning. In this paper, we classify gestures using data collected from a curved piezoelectric sensor, including elbow movement, wrist turning, bending, coughing, and neck bending. The classification process relies on sensor. Machine learning algorithms enabled with K-mer are developed optimized to perform gesture (HGR) acquired achieve best results. Three algorithms, namely support vector (SVM), random forest (RF), k-nearest neighbor (k-NN), performed analyzed K-mer. input parameters such as subsequence length (K), number cuts, penalty parameter (C), trees (n_estimators), maximum depth tree (max_depth), nearest neighbors (k) for three modified accuracy. proposed model was evaluated its accuracy percentage, recall score, precision F-score value. We promising results 94.11 ± 0.3%, 97.18 0.4%, 96.90 0.5% SVM, RF, k-NN, respectively. execution time run program optimal is 19.395 1 s, 5.941 3.832 s