WiFi-Based Human Activity Recognition Systems

WiFi-based human activity recognition systems exploit the inherent sensitivity of wireless signals to human motion, posture and presence. As individuals move within an indoor environment, their bodies perturb multipath reflections of commercial WiFi signals, producing characteristic variations in received signal strength and channel state information. By analysing these fluctuations through signal‐processing pipelines and machine‐learning models, it is possible to infer activities such as walking, sitting, falling or gesturing without requiring wearable devices or cameras. These non-intrusive systems draw on advances in feature extraction, pattern classification and deep neural architectures to achieve robust recognition even in complex, non-line-of-sight settings. Their global significance extends across smart homes, eldercare monitoring, security surveillance and human–computer interaction, offering cost-effective, privacy-preserving and scalable solutions that leverage the ubiquity of existing WiFi infrastructure.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Recent surveys of multimodal indoor monitoring have highlighted the integration of WiFi sensing with complementary modalities such as video and inertial data to bolster recognition accuracy and contextual understanding. These reviews map the latest feature-fusion strategies and benchmark datasets designed for elderly-care applications, illustrating how combining channel state information with image and accelerometer streams can mitigate individual sensor limitations. Concurrently, developments in few-shot learning have been applied to channel state information platforms, enabling trained models to generalise rapidly to new environments or unencountered activities with minimal additional data. This line of work leverages transfer, metric and meta-learning frameworks to address domain shift and reduce the burden of extensive retraining. In the healthcare domain, contactless WiFi monitoring systems have matured to detect critical events—such as falls, sleep disturbances and respiratory anomalies—through refined signal-processing chains and lightweight convolutional neural network classifiers. These systems demonstrate real-world viability by achieving high sensitivity and specificity in pilot deployments, underscoring the promise of WiFi-based sensing to deliver unobtrusive, continuous health surveillance without compromising user comfort or privacy.

WiFi-Based Human Activity Recognition Systems publication trend

The graph below shows the total number of articles in wifi-based human activity recognition systems across all publications each year (not limited to Nature Index journals).

Technical terms

Channel State Information (CSI): Fine-grained measure of the amplitude and phase responses across individual WiFi subcarriers, used to capture subtle environmental changes induced by human movement.

Received Signal Strength Indicator (RSSI): Coarse metric of the power level received by a WiFi device, reflecting aggregate signal attenuation and multipath effects.

Few-shot Learning: Machine-learning approach that enables models to adapt to new classes or settings using only a small number of labelled examples.

Convolutional Neural Network (CNN): Deep-learning architecture that applies convolutional filters to structured data (such as time-series or images) for hierarchical feature learning and classification.

References

1. Non-contact multimodal indoor human monitoring systems: A survey. Information Fusion (2024).
2. Review of few-shot learning application in CSI human sensing. Artificial Intelligence Review (2024).
3. Contactless WiFi Sensing and Monitoring for Future Healthcare - Emerging Trends, Challenges, and Opportunities. IEEE Reviews in Biomedical Engineering (2023).