Human Activity Classification Based on Data Analysis and Feature Extraction

Abstract

Human behavior recognition is one of the most important research directions in the field of computer vision, and it plays an important role in the fields of rehabilitative medicine, auxiliary security, and scene entertainment. To address the shortcomings of traditional HAR recognition methods with tedious feature extraction and severe overfitting, we propose a human behavior recognition model based on XGBoost and feature simplification methods with a limited data set. The model uses the XGBoost algorithm to classify the collected sensor data to recognize human behaviors. In addition, to improve the efficiency and accuracy of the model, we also propose a feature simplification method to reduce the computational complexity and the risk of model overfitting by reducing the number of features. Experimental results show that the model has high accuracy and computational efficiency and can be applied to different human behavior recognition scenarios. CCS Concepts: Computing methodologies∼Machine learning∼Machine learning approaches