Human Activity Recognition with a 6.5 GHz Reconfigurable Intelligent Surface for Wi-Fi 6E

Abstract: Human Activity Recognition (HAR) is the identification and classification of static and dynamic human activities, which find applicability in domains like healthcare, entertainment, security, and cyber-physical systems. Traditional HAR approaches rely on wearable sensors, vision-based systems, or ambient sensing, each with inherent limitations such as privacy concerns or restricted sensing conditions. Recently, Radio Frequency (RF)-based HAR has emerged, relying on the interaction of RF signals with people to infer activities. Reconfigurable Intelligent Surfaces (RISs) offers significant potential in this domain by enabling dynamic control over the wireless environment, thus enhancing the information extracted from RF signals. We present an Hand Gesture Recognition (HGR) approach that employs our own 6.5 GHz RIS design to manipulate the RF medium in an area of interest. We validate the capability of our RIS to control the medium by characterizing its steering response, and further we gather and publish a dataset for HGR classification for three different hand gestures. By employing two Convolutional Neural Networks (CNNs) models trained on data gathered under random and optimized RIS configuration sequences, we achieved classification accuracies exceeding 90%.

• The paper proposes a novel RIS design for the 6.5 GHz band that enables enhanced human gesture recognition in Wi-Fi 6E environments.
• It employs optimized data acquisition via FCAO and CNN-based methods, significantly improving gesture classification accuracy.
• Experimental tests in anechoic chambers validate effective beamforming, paving the way for scalable RIS improvements in RF sensing.

Insights into Human Activity Recognition Using Reconfigurable Intelligent Surfaces

The paper presented provides a comprehensive examination of Human Activity Recognition (HAR) using Reconfigurable Intelligent Surfaces (RIS) at a 6.5 GHz frequency, which is relevant to the Wi-Fi 6E standard. This work addresses some of the intrinsic challenges associated with traditional HAR methods, such as privacy issues or limited sensing environments, by leveraging RF-based HAR systems augmented with RIS technology to dynamically influence the wireless medium.

The authors focus on employing a RIS designed for a specific frequency that aligns with Wi-Fi 6E to achieve Human Gesture Recognition (HGR), specifically targeting hand gestures. This decision is informed by the dimensions of their RIS, which spatially limit the sensing capabilities largely to hand movements. The RIS, comprised of 64 unit cells controlled by PIN diodes with 1-bit phase control, showcases an enhanced ability to manipulate the phase and beamforming of the RF signals, proving instrumental in optimizing the information gathered from the wireless environment.

The paper articulates several noteworthy contributions:

1. Design and Fabrication of the RIS: The team developed a unit cell tailored for the 6.5 GHz spectrum, crafted from a carefully chosen substrate material and implemented with PIN diode control to achieve desired phase shifts.
2. Characterization Techniques: The RIS was characterized through anechoic chamber experiments which validated its capacity to adjust beam directionality. Measurements revealed effective beamforming especially at some deviation from the targeted frequency, indicating areas for potential refinement in future iterations of the hardware.
3. Optimized Data Acquisition Methods: By leveraging an adaptation of the Frame Configuration Alternating Optimization (FCAO) technique, the authors sought to minimize mutual coherence in the RIS configurations, enhancing the discrimination ability of the system when identifying gestures. It was demonstrated that optimized configurations significantly improved gesture classification accuracy when limited frequency data points were used.
4. Broadband Sensing Capabilities: Beyond techniques focusing on singular narrowband frequencies, the study also explored broadband sensing across a 5.0 GHz to 6.5 GHz range. This rich dataset was processed by convolutional neural networks (CNNs) to achieve high classification accuracy.
5. CNN-based Classification: Two distinct CNN architectures were assessed. The first method emphasized narrowband, configuration-critical model inputs where optimization through FCAO had a notable positive impact. In contrast, the second model, exploiting richer spectral data, inherently provided competitive accuracies with both random and optimized configurations, thus challenging the necessity of elaborate optimization in broadband scenarios.

Overall, the implications of the study suggest several promising avenues for future research. The potential to scale up RIS designs suggests applications beyond current Wi-Fi standards, and publication of the dataset can stimulate further research initiatives in the RF sensing community. The convergence of RF-based HAR with 6G networks accentuates the importance of RIS technology in augmenting traditional sensor-based methods, providing a path toward robust and non-invasive human-computer interaction systems. The alignment of this research with real-world constraints, such as the energy budgets of RIS tiles for ubiquitous Wi-Fi, positions it as a high-impact contribution in the evolving field of intelligent atmospheric radio manipulation.