[Paper Review] From Real to Complex: Enhancing Radio-based Activity Recognition Using Complex-Valued CSI

Bo Wei, Wen Hu+2 more|arXiv (Cornell University)|Apr 25, 2018

Indoor and Outdoor Localization Technologies40 references6 citations

TL;DR

This paper proposes a novel device-free activity recognition system using complex-valued Channel State Information (CSI) to enhance robustness against radio frequency interference (RFI). By leveraging sparse representation classification (SRC) on complex-valued CSI, the method improves recognition accuracy by up to 10% in RFI-degraded environments, outperforming real-valued CSI approaches and demonstrating feasibility with minimal hardware setup.

ABSTRACT

Activity recognition is an important component of many pervasive computing applications. Radio-based activity recognition has the advantage that it does not have the privacy concern and the subjects do not have to carry a device on them. Recently, it has been shown channel state information (CSI) can be used for activity recognition in a device-free setting. With the proliferation of wireless devices, it is important to understand how radio frequency interference (RFI) can impact on pervasive computing applications. In this paper, we investigate the impact of RFI on device-free CSI-based location-oriented activity recognition. We present data to show that RFI can have a significant impact on the CSI vectors. In the absence of RFI, different activities give rise to different CSI vectors that can be differentiated visually. However, in the presence of RFI, the CSI vectors become much noisier and activity recognition also becomes harder. Our extensive experiments show that the performance of state-of-the-art classification methods may degrade significantly with RFI. We then propose a number of counter measures to mitigate the impact of RFI and improve the location-oriented activity recognition performance. We are also the first to use complex-valued CSI to improve the performance in the environment with RFI.

Motivation & Objective

• To address the critical challenge of radio frequency interference (RFI) in device-free CSI-based activity recognition systems.
• To improve the robustness and accuracy of location-oriented activity recognition in dynamic, real-world environments with high RFI levels.
• To investigate the feasibility and advantages of using complex-valued CSI instead of real-valued CSI for enhanced signal representation and classification performance.
• To develop and evaluate a sparse representation classification (SRC) framework tailored for complex-valued CSI under RFI conditions.
• To demonstrate system feasibility using low-cost, embedded wideband radio platforms with minimal hardware requirements.

Proposed method

• The system uses complex-valued CSI obtained from off-the-shelf WiFi chipsets (e.g., Intel 5300, Atheros 9390) as input for activity recognition.
• It applies sparse representation classification (SRC) based on ℓ1-optimization to model and classify CSI vectors, leveraging the sparsity of activity patterns in a dictionary of known activities.
• The method treats each activity as a sparse linear combination of basis vectors from a trained dictionary, enabling robust classification even under noisy or interfered CSI.
• Complex-valued CSI is used to preserve phase and amplitude information, enhancing discriminative power compared to real-valued magnitude-only representations.
• The system is evaluated using an embedded WASP platform to emulate various wireless protocols and bandwidths (e.g., 20 MHz), simulating real-world deployment scenarios.
• Countermeasures against RFI include signal preprocessing and the use of SRC’s inherent noise resilience, with performance validated across multiple interference conditions.

Experimental results

Research questions

• RQ1 How does radio frequency interference (RFI) degrade the performance of CSI-based device-free activity recognition systems?
• RQ2 Can complex-valued CSI improve classification accuracy compared to real-valued CSI in the presence of RFI?
• RQ3 To what extent can sparse representation classification (SRC) enhance robustness to RFI in CSI-based activity recognition?
• RQ4 What is the performance gain of using complex-valued CSI with SRC in real-world, interference-prone environments?
• RQ5 How scalable and practical is the proposed system with minimal hardware (e.g., one transmitter-receiver pair) in real indoor settings?

Key findings

• RFI significantly degrades the quality of CSI vectors, making activity recognition more difficult and reducing performance of state-of-the-art classifiers.
• Complex-valued CSI preserves phase and amplitude information, enabling better discrimination between activities compared to real-valued magnitude-only CSI.
• The proposed SRC-based method using complex-valued CSI improves recognition accuracy by up to 10% in RFI-affected environments compared to real-valued CSI approaches.
• SRC demonstrates strong robustness to noise and interference, outperforming traditional methods like SVM and k-NN in RFI conditions.
• The system achieves reliable performance with only one pair of transceivers in a one-bedroom apartment, enabling easy deployment and maintenance.
• The prototype demonstrates feasibility across different wireless bandwidths (e.g., 20 MHz), showing that high accuracy is achievable even with limited spectrum.