Human Activity Recognition via Ultra-Wideband Data: A Framework for Dimensionality Reduction, Pattern Discovery, and Predictive Modeling

Nahid Sahel Gozin; Reza Sedaghat; Prathap Siddavaatam

cs.LGSep 25, 2026

Human Activity Recognition via Ultra-Wideband Data: A Framework for Dimensionality Reduction, Pattern Discovery, and Predictive Modeling

Authors: Nahid Sahel Gozin, Reza Sedaghat, Prathap Siddavaatam

Organizations: OPR-AL Lab, Toronto Metropolitan University, 350 Victoria Street, Toronto, ON M5B 2K3, Canada

Abstract

Recent advances in sensor technology have enabled more effective human activity recognition (HAR), particularly in real-time systems with limited computational resources. However, Ultra-Wideband (UWB) radar data remain challenging due to high dimensionality, noise, complexity, and nonlinear characteristics. This research proposes a framework to efficiently reduce data size, uncover significant patterns, and classify six activity types (Standff, Liedown, Noactivity, Sit, Stand, and Walk) from UWB signals with high accuracy. Two novel dimensionality reduction techniques are introduced in this paper. The first, Clustered Polynomial Expansion with Incremental PCA (CPE-IPCA), combines clustering and polynomial feature expansion with Incremental PCA, preserving 100% of the variance in only 50 components. The second, Post-PCA Standardization Approach (PPSA), standardizes data after PCA and retains 99.1% of the variance in 80 components, achieving superior compression and computational efficiency compared to conventional nonlinear methods. Frequent patterns are identified using Apriori and FP-Growth, which are then classified with Random Forest and a Vector Space Model (VSM). The framework achieves 100% accuracy with Random Forest on CPE-IPCA and 99% on PPSA, while VSM attains 100% precision, recall, and F1 on PPSA and near-perfect performance on CPE-IPCA (precision 1.00, recall 0.98-1.00, F1 0.99-1.00), demonstrating a fast, interpretable, and robust HAR system suitable for healthcare, assisted living, and smart environments.

Figures & tables

Activity Class	Number of Instances
No Activity	7,032
Lie Down	1,729
Standff	1,668
Sit	2,230
Stand	2,236
Walk	1,898

TABLE I: Activity Class Distribution

Component	Explained Variance (%)	Cumulative Variance (%)
PC1	12.98	12.98
PC2	10.34	23.32
PC3	9.02	32.34
PC4	8.51	40.85
PC5	7.43	48.28
PC6	6.81	55.09

TABLE II: Explained and Cumulative Variance by CPE-IPCA (Top 50 Components)

Component	Explained Variance (%)	Cumulative Variance (%)
PC1	17.48	17.48
PC2	17.24	34.72
PC3	9.61	44.33
PC4	9.46	53.79
PC5	4.66	58.45
PC6	4.55	63.00

TABLE III: Explained and Cumulative Variance: Post-PCA Standardization

Panel of Figure 4 — Fig. 1: Explained variance (left) and cumulative explained variance (right) for each principal component after PCA. PC1 captures 4.73% variance; 723 components are needed to retain 90% of the variance.

Panel of Figure 5 — Fig. 2: The CPE-IPCA analysis identified explained and cumulative explained variances for each principal component. PC1 explains 13% of the variance, while PC13 marks the elbow with 86% cumulative variance.

Panel of Figure 6 — Fig. 3: PPSA: The left plot shows the individual explained variance for each principal component, highlighting PC1 with the highest contribution of 17.5%. The right plot illustrates the cumulative explained variance, which reaches 99.1%.

Rule Set	Rule Count	Mean	Std Dev
Confidence 0.1	39,851 $\times$ 10	0.254	0.110
Confidence 0.3	16,340 $\times$ 10	0.359	0.045
Confidence 0.5	26 $\times$ 10	0.516	0.018
Lift (all)	45,150 $\times$ 10	1.086	0.122
Support 0.1	376 $\times$ 10	0.112	0.011
Support 0.3–0.5	0	—	—

TABLE IV: Post-PCA: Summary of Rule Counts and Key Metrics

Panel of Figure 8 — Fig. 4: Top 20 association rules ranked by lift.

Panel of Figure 9 — Fig. 5: Top 20 association rules ranked by Support.

Panel of Figure 10 — Fig. 6: PPSA: Network graphs from Apriori (top) and FP-Growth (bottom) showing principal component associations. Edge colors indicate lift strength; FP-Growth offers faster detection.

Panel of Figure 11 — Fig. 7: Top 20 association rules ranked by lift.

Panel of Figure 12 — Fig. 8: Top 20 association rules ranked by Support.

Filtering Type	Threshold	Rule Set Size	Mean Value
Confidence-based	0.05	59,022	0.22
Confidence-based	0.2	30,797	0.33
Confidence-based	0.4	4,770	0.46
Lift (all)	0.05, 0.2, 0.4	59,166	1.42
Support-based	0.05	2,218	0.07
Support-based	0.2	0	—

TABLE V: CPE-IPCA: Rule Summary Across Thresholds

Panel of Figure 14 — Fig. 9: CPE-IPCA: Network graphs from Apriori (top) and FP-Growth (bottom) showing principal component associations. Edge colors denote lift strength; FP-Growth detects faster.

Class	PPSA				CPE-IPCA
Class	Prec.	Rec.	F1	Sup.	Prec.	Rec.	F1	Sup.
Standff	1.00	1.00	1.00	11021	1.00	1.00	1.00	16558
Liedown	1.00	0.20	0.33	5	1.00	0.88	0.94	17
Noactivity	0.98	0.99	0.99	2460	0.97	0.97	0.97	1128
Sit	1.00	1.00	1.00	22	1.00	1.00	1.00	11
Stand	1.00	0.80	0.89	25	1.00	1.00	1.00	26

TABLE VI: Random Forest on PPSA and CPE-IPCA rule datasets (30% test) with Precision, Recall, F1, Support, accuracy, and class averages.

Actual \Predicted	Standff	Liedown	Noactivity	Sit	Stand	Walk
Standff	1.00	0.00	0.00	0.00	0.00	0.00
Liedown	0.80	0.20	0.00	0.00	0.00	0.00
Noactivity	0.01	0.00	0.99	0.00	0.00	0.00
Sit	0.00	0.00	0.00	1.00	0.00	0.00
Stand	0.20	0.00	0.00	0.00	0.80	0.00
Walk	0.00	0.00	0.00	0.00	0.00	1.00

TABLE VII: Random Forest Confusion Matrix on Activity Rule-Set Extracted from PPSA

Actual \Predicted	Standff	Liedown	Noactivity	Sit	Stand	Walk
Standff	1.00	0.00	0.00	0.00	0.00	0.00
Liedown	0.06	0.88	0.06	0.00	0.00	0.00
Noactivity	0.03	0.00	0.97	0.00	0.00	0.00
Sit	0.00	0.00	0.00	1.00	0.00	0.00
Stand	0.00	0.00	0.00	0.00	1.00	0.00
Walk	0.00	0.00	0.00	0.00	0.00	1.00

TABLE VIII: Random Forest Confusion Matrix on Activity Rule-Set Extracted from CPE-IPCA

Class	PPSA				CPE-IPCA
Class	Prec.	Rec.	F1	Sup.	Prec.	Rec.	F1	Sup.
Standff	1.00	1.00	1.00	11094	1.00	1.00	1.00	16558
Liedown	1.00	1.00	1.00	9	1.00	0.99	1.00	17
Noactivity	1.00	1.00	1.00	2388	1.00	0.98	0.99	1128
Sit	1.00	1.00	1.00	18	1.00	1.00	1.00	11
Stand	1.00	1.00	1.00	27	1.00	1.00	1.00	26

TABLE IX: VSM performance on PPSA and CPE-IPCA rule features (30% test) with Precision, Recall, F1, Support, and overall accuracy per activity class.

Actual $\backslash$ Pred	Standff	Liedown	Noactivity	Sit	Stand	Walk
Standff	1	0	0	0	0	0
Liedown	0	1	0	0	0	0
Noactivity	0	0	1	0	0	0
Sit	0	0	0	1	0	0
Stand	0	0	0	0	1	0
Walk	0	0	0	0	0	1

TABLE X: VSM Confusion Matrix on Activity Rule-Set Extracted from PPSA

Actual $\backslash$ Pred	Standff	Liedown	Noactivity	Sit	Stand	Walk
Standff	1	0	0	0	0	0
Liedown	0	1	0	0	0	0
Noactivity	0.02	0	0.98	0	0	0
Sit	0	0	0	1	0	0
Stand	0	0	0	0	1	0
Walk	0	0	0	0	0	1

TABLE XI: VSM Confusion Matrix on Activity Rule-Set Extracted from CPE-IPCA

Explore similar work

WISE-HAR: A Generalizable Ensemble Deep Learning Framework for WiFi-Based Human Activity Recognition
Jun 2, 2026Maheen Arshad, Qindeel E Zahra, Muhammad Khuram ShahzadHuman Activity Recognition Wi-Fi Sensing
Department of Computing, School of Electrical Engineering and Computer Science National University of Sciences and Technology (NUST), Islamabad, Pakistan
IBIS: A Hybrid Inception-BiLSTM and SVM Ensemble for Robust Doppler-based Human Activity Recognition
Oct 28, 2025Alison M. Fernandes, Hermes I. Del Monego, Bruno S. Chang +3Human Activity Recognition Domain Generalization
Universidade Tecnol´ogica Federal do Paran´a (UTFPR), Programa de P´os-Graduac¸ ˜ao em Engenharia El´etrica e Inform´atica Industrial (CPGEI-CT), Av. Sete de Setembro, 3165, Curitiba, 80230-901, Brazil · INESC TEC, Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias, Porto, 4200-465, Porto, Portugal
LITEWAY: LIghtweight HAR via Temporal Efficient highWAY
Aug 10, 2026Dominique Nshimyimana, Vitor Fortes Rey, Mengxi Liu +2Efficient Neural Network Inference Human Activity Recognition
RPTU, DFKI Kaiserslautern, Germany

Jun 2, 2026J/K move · Enter open · S save

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

Maheen Arshad, Qindeel E Zahra, Muhammad Khuram Shahzad

Department of Computing, School of Electrical Engineering and Computer Science National University of Sciences and Technology (NUST), Islamabad, Pakistan

Human Activity Recognition Wi-Fi Sensing

Open paper AskPDF arXiv

Human Activity Recognition (HAR) using WiFi signals has emerged as a transformative technology for smart homes, healthcare monitoring, security systems, and ambient assisted living. Unlike traditional camera-based systems that raise significant privacy concerns and fail in low-light conditions, or wearable sensors that require user compliance, WiFi-based HAR is non-intrusive, privacy-preserving, cost-effective, and works seamlessly in any lighting condition. This paper presents a comprehensive approach to recognize three distinct human activities: "No Presence" (empty room), "Walking", and "Walking + Arm-waving" using the Wallhack1.8k WiFi spectrogram dataset. We propose three key improvements to address the main challenges in WiFi-based HAR. First, to address high performance variance, we implement ensemble learning with five different CNN architectures (Deep CNN, Wide CNN, MobileNetV2, ResNet50V2, and EfficientNetB0). Second, to address the small dataset size limitation, we apply aggressive data augmentation techniques including time-warping, frequency masking, and noise addition. Third, to evaluate real-world generalization capability, we perform cross-scenario evaluation (training on Line-of-Sight and testing on Non-Line-of-Sight) and cross-antenna evaluation (training on Biquad antenna and testing on PIFA antenna). Our ensemble model achieved a test accuracy of 94.87% on the LOS scenario with Biquad antenna, outperforming the best individual model by 0.66%. Data augmentation improved Random Forest performance from 60% to 95%. Cross-scenario evaluation showed minimal accuracy drops of only 1.37% and 2.07%, demonstrating strong generalization capabilities. The results indicate that the proposed approach is robust, reliable, and suitable for real-world deployment in diverse environments with different hardware configurations.

Abstract

Explore similar work

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

IBIS: A Hybrid Inception-BiLSTM and SVM Ensemble for Robust Doppler-based Human Activity Recognition

LITEWAY: LIghtweight HAR via Temporal Efficient highWAY