A CONFIDENCE CALIBRATED LIGHTWEIGHT FUSION MODEL FOR ON DEVICE HUMAN ACTIVITY RECOGNITION
ICTACT Journal on Soft Computing · 01.10.2026 · 17 (3) · 4365-4372
DOI: 10.21917/ijsc.2026.0600 ↗Kurz gefasst
Fachartikel in der Zeitschrift ICTACT Journal on Soft Computing, online erschienen am 01.10.2026. Verfasst von Kamal Narayan Budholia und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Sensor-based human activity recognition (HAR) is often deployed under subject shift, uncertain predictions, and tight resource budgets. We introduce CALM- HAR, a confidence-calibrated, lightweight fusion framework designed for reliable on-device inference. CALM-HAR couples a tiny temporal encoder (shallow 1-D CNN) with tree-based learners (Random Forest, XGBoost) and applies post-hoc temperature scaling to obtain well-calibrated class probabilities. A confidence-weighted late-fusion head combines learners in a data-dependent manner, using predictive entropy to modulate each model’s contribution without adding heavy meta-networks. We evaluate under subject-independent (leave-one-subject-out) and cross-dataset (train on UCI HAR, test on PAMAP2) protocols and report accuracy, macro-Precision/Recall/F1 together with Expected Calibration Error (ECE), negative log-likelihood (NLL), and CPU-bound latency/size. Across settings, CALM-HAR maintains competitive macro-F1 while consistently improving probability calibration relative to un-calibrated stacking and single-model baselines, yielding dependable decisions at low cost. The paper details pre-processing, hyper parameters and evaluation settings to support transparent replication. These results indicate that careful calibration and confidence-aware fusion offer a practical alternative to heavier architectures for robust, edge-oriented HAR.
Zitieren
APA
Budholia, K. N., & Pandey, A. (2026). A CONFIDENCE CALIBRATED LIGHTWEIGHT FUSION MODEL FOR ON DEVICE HUMAN ACTIVITY RECOGNITION. ICTACT Journal on Soft Computing, 17(3), 4365-4372. https://doi.org/10.21917/ijsc.2026.0600
Vancouver
Budholia KN, Pandey A. A CONFIDENCE CALIBRATED LIGHTWEIGHT FUSION MODEL FOR ON DEVICE HUMAN ACTIVITY RECOGNITION. ICTACT Journal on Soft Computing. 2026;17(3):4365-4372. doi:10.21917/ijsc.2026.0600
BibTeX
@article{Budholia2026,
author = {Budholia, Kamal Narayan and Pandey, Ankur},
title = {A CONFIDENCE CALIBRATED LIGHTWEIGHT FUSION MODEL FOR ON DEVICE HUMAN ACTIVITY RECOGNITION},
journal = {ICTACT Journal on Soft Computing},
year = {2026},
volume = {17},
number = {3},
pages = {4365--4372},
doi = {10.21917/ijsc.2026.0600}
}
EUVIMED