A Domain-Independent Generative Adversarial Network for Activity Recognition Using WiFi CSI Data

Abstract

Over the past years, device-free sensing has received considerable attention due to its unobtrusiveness. In this regard, context recognition using WiFi Channel State Information (CSI) data has gained popularity, and various techniques have been proposed that combine unobtrusive sensing and deep learning to accurately detect various contexts ranging from human activities to gestures. However, research has shown that the performance of these techniques significantly degrades due to change in various factors including sensing environment, data collection configuration, diversity of target subjects, and target learning task (e.g., activities, gestures, emotions, vital signs). This problem, generally known as the domain change problem, is typically addressed by collecting more data and learning the data distribution that covers multiple factors impacting the performance. However, activity recognition data collection is a very labor-intensive and time consuming task, and there are too many known and unknown factors impacting WiFi CSI signals. In this paper, we propose a domain-independent generative adversarial network for WiFi CSI based activity recognition in combination with a simplified data pre-processing module. Our evaluation results show superiority of our proposed approach compared to the state of the art in terms of increased robustness against domain change, higher accuracy of activity recognition, and reduced model complexity.

MetaGanFi

“…DMNet [19] and GNG [15] also utilized GAN to produce additional samples and facilitate activity recognition. A recent work [39] explored GAN to learn domain independent features for activity recognition. These prior works in essence exploit the data generation ability of GAN model to synthesize virtual samples in target domains and facilitate domain adaption.…”

Sensing the care: advancing unobtrusive sensing solutions to support informal caregivers of older adults with cognitive impairment

“…In the later or pre-implementation stage (in real-life settings), actors or PwDs can be invited for determinative performance analysis of the system. Along with that, this dataset opens new avenues for gaining insights into the ubiquitousness and domain-transfer applications of CSI-based sensing as it contains parameters like different activity locations, participant characteristics, and placement of Tx-Rx [211]. Insights into the ubiquitousness can help in making informed decisions on using these advanced sensing systems.…”

“…In a recent study, the problem of domain change due to environment and person was identified, and to mitigate that a domain-independent generative adversarial network was proposed and evaluated for gesture recognition [224].…”

“…The available research demonstrates the use of CSI for recognizing physical activity such as falling, walking, sitting, and hand gestures; physiological activity such as heart and breathing rate; and behaviors such as sleep patterns, eating-drinking habits, and physical fitness [220] [221] [222]. Though the achieved success of CSI in HAR is fascinating, its inability to generalize across domains due to data mismatch hinders its real-world implementation [223] [224]. Data mismatch occurs when there is a significant difference between the distributions of trained and testing data.…”

“…For example, the work by Zinys et al [223] proposes the use of GAN-based architecture to improve performance when different environments or persons data is used for training and testing. In another study, a novel approach using a wavelet template and stretch-limited dynamic time wrapping was proposed to overcome the impact of distance and orientation on the performance of CSI-based HAR [224]. Since the majority of available studies were conducted in labs or in controlled setups, a clear understanding of the domains causing data mismatch (individually or when combined) as expected in CSI-based HAR in real-world scenarios cannot be drawn.…”