WISE-HAR: A Generalizable Ensemble Deep Learning Framework for WiFi-Based Human Activity Recognition

2026-06-02 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Internet of Things (IoT) & Connected Devices · Depth: Expert, quick

Summary

WISE-HAR, a generalizable ensemble deep learning framework, addresses challenges in WiFi-based Human Activity Recognition (HAR), a non-intrusive, privacy-preserving, and cost-effective alternative to camera or wearable systems. The framework recognizes "No Presence", "Walking", and "Walking + Arm-waving" using the Wallhack1.8k WiFi spectrogram dataset. It incorporates three key improvements: ensemble learning with five CNN architectures (Deep CNN, Wide CNN, MobileNetV2, ResNet50V2, EfficientNetB0) to reduce performance variance; aggressive data augmentation (time-warping, frequency masking, noise addition) to overcome small dataset limitations; and rigorous cross-scenario (Line-of-Sight/Non-Line-of-Sight) and cross-antenna (Biquad/PIFA) evaluation for real-world generalization. The ensemble model achieved 94.87% test accuracy on the LOS scenario with Biquad antenna, outperforming individual models by 0.66%, and data augmentation boosted Random Forest performance from 60% to 95%. Cross-scenario evaluation showed minimal accuracy drops of 1.37% and 2.07%, confirming its robustness for diverse deployments.

Key takeaway

For Machine Learning Engineers developing HAR systems, WISE-HAR's approach offers a blueprint for achieving high accuracy and generalization in non-intrusive sensing. You should consider integrating ensemble learning with diverse CNN architectures and aggressive data augmentation techniques like time-warping and frequency masking. This strategy will enhance your model's robustness and reliability, ensuring strong performance across varied real-world deployment scenarios and hardware configurations, minimizing accuracy drops in new environments.

Key insights

Ensemble deep learning with data augmentation significantly enhances WiFi-based HAR generalization and robustness across diverse environments.

Principles

• Ensemble learning mitigates performance variance in HAR.
• Data augmentation is crucial for small HAR datasets.
• Cross-scenario testing validates real-world HAR generalization.

Method

Combine five CNN architectures via ensemble learning, apply time-warping, frequency masking, and noise addition for data augmentation, then evaluate using cross-scenario and cross-antenna tests.

In practice

• Implement ensemble CNNs for robust HAR.
• Use time-warping for WiFi spectrogram augmentation.
• Test HAR models across different antenna types.

Topics

• WiFi Human Activity Recognition
• Ensemble Deep Learning
• Data Augmentation
• Model Generalization
• CNN Architectures
• Non-intrusive Sensing

Best for: Research Scientist, AI Scientist, Machine Learning Engineer

Related on AIssential

• KAN-MLP-Mixer: A comprehensive investigation of the usage of Kolmogorov-Arnold Networks (KANs) for improving IMU-based Human Activity Recognition — Strategically combining KANs for data embedding and classification with MLPs for feature mixing significantly enhances noisy IMU-based HAR.
• Generalization and Membership Inference Attack a Practical Perspective — Enhanced model generalization through augmentation and early stopping significantly reduces Membership Inference Attack success.
• Physics-Guided Attention in a Lightweight TCN for Efficient WiFi CSI-Based Human Activity Recognition — Incorporating physics-guided inductive biases into lightweight TCNs significantly improves WiFi CSI-based HAR efficiency and performance.
• Generative Data Augmentation for Skeleton Action Recognition — A conditional generative pipeline synthesizes diverse skeleton data to augment action recognition, improving performance in low-data settings.
• Ensembles of Ensembles of Ensembles: A Guide to Stacking — Ensembling diverse models across multiple layers enhances predictive performance and robustness for tabular and time series data.
• Complex & Intelligent Systems, Volume 12, Issue 4, April 2026 — The issue highlights diverse AI advancements across networking, computer vision, NLP, and secure systems.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.