Physical Human Activity Recognition Based on Empirical Model Decomposition Coupled with Classification Models

Due to the rapid increase in data collected through wearable devices and smartphones a high demand for developing human activity recognition techniques has emerged. However, human activity data is a complex and difficult task to identify due to its varied sources of data. In response, this paper proposes an innovative transformation technique for HAR. This proposed methodology comprises three key phases: feature extraction, feature selection, and features classification using an ensemble model. Empirical mode decomposition receives the HAR data (EMD).Different modes are extracted and analyzed. A set of features will be extracted from each position. Several machine learning models receive the extracted features.We evaluated the proposed method using public dataset, the results obtained demonstrated that the proposed approach surpasses existing HAR models.