TY  - EJOU
AU  - Sankaran, Ashwin 

TI  - Intra-Modal Feature Translation for Robust Kinematic Modeling in Wearable Health Applications
T2  - Journal on Artificial Intelligence

PY  - 2026
VL  - 8
IS  - 1
SN  - 2579-003X

AB  - Human Motion Kinetics Recognition (HMKR) plays a pivotal role in advancing intelligent monitoring systems for public safety, machine interaction, and elderly-care technologies. Despite the increasing availability of affordable Inertial Measurement Units (IMUs), achieving accurate full-body motion analysis remains challenging due to sensor noise and variability across real-world environments. Addressing these limitations requires robust processing strategies capable of handling complex inertial time-series data. This research introduces an optimized HMKR system that leverages a comprehensive pipeline focusing on inertial sensor time-series data. In this research, we introduce an HMK-based recognition system that primarily focuses on six-degree-of-freedom (6-DoF) sensors that capture translational and rotational movements of the human body to improve health and safety applications. The study also performed a simulation study on two public benchmark datasets. The research’s main contribution lies in integrating a preprocessing approach, including state-of-the-art denoising and a multi-stream feature set (statistical, frequency-domain, and vector features), that effectively mitigates sensor limitations and high-dimensionality issues via principal component analysis. These results justify the proposed system’s reliability and its significance to substantially enhance health and safety monitoring applications, particularly in unconstrained public health contexts.
KW  - Feature space; human motion analysis; machine learning; classification; movement analysis; categorical boosting

DO  - 10.32604/jai.2026.080854