Towards improved wireless communication and sensing with CSI augmentation

• Main Author: Zhang, Yanbo
• Other Authors: Mo Li
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/164889
• Institution: Nanyang Technological University

Wireless communication technology has become one key technology that supports the operation of our society and promotes economy growth. With the ubiquitous deployment of wireless devices, wireless signal based sensing techniques have become a hot research topic in recent years. For both communication and sensing systems, the awareness of wireless channel is very important. Both academia and industry adopt channel state information(CSI) as a metric that quantifies the channel quality. Despite its significance for wireless systems, CSI obtained with a hardware system is usually limited in its spatial diversity, time resolution and accuracy. We propose CSI augmentation in terms of the three aspects, based on which we completed three research projects. We investigate how the expansion of array size may improve the spatial diversity of the CSI obtained with state-of-the-art Wi-Fi system and increase its throughput. With comprehensive Wi-Fi measurement studies with augmented antennas, we identify the potential performance gain atop spatial diversity gains from existing technologies like MIMO and beamforming. We propose a general Wi-Fi intelligent antenna selection scheme with full system implementation that can be easily integrated with commodity Wi-Fi AP. The proposed system provides substantially improved throughput for downlink traffics. Our experimental evaluation suggests that our design improves Wi-Fi throughput up to 1.56x, and 1.47x in average, in real user-based evaluation. Vision based face recognition has been widely adopted for diverse purposes but is known to be inaccurate with challenging environment conditions, such as foggy or smoky weather, poor lighting, and blockage by objects like facial mask. We invent an acoustic based facial recognition system that operates atop commercial devices. We propose acoustic facial spectrogram – an acoustic facial fingerprint for describing human facial spatial characteristics. Two special challenges are faced when devising accurate and robust face recognition. First, with commercial hardware, there only exists limited noise-free acoustic frequency band, which significantly limits the frequency diversity in channel estimation, and as a result limits the time resolution of the derived facial spectrogram. Second, when wearing mask, a great portion of facial features are blocked, which may lead to inaccurate facial profiling. This paper proposes two novel techniques to address the above two challenges. A prototype system is built with inexpensive commercial devices. Extensive experimental results demonstrate that the proposed system achieves over 93% average recognition accuracy for cases with and without mask blockage. Device-free hand-writing systems identify the content that a user writes by hand movement in the air, thus providing an intuitive human computer interface. We propose a Wi-Fi handwriting recognition system built with commodity Wi-Fi APs. Unlike most existing machine learning based hand-writing recognition systems, which are often subject to severe limitations in generality, e.g., high training overhead when adapted across hand-writing alphabets, environments, and users, our proposed system is designed with unique consideration of its generality when applied to practice – being application-transferable, environment-agnostic, and user-independent. With little training overhead, the system behaves inclusively to different users, environments, and applications, stemming from a comprehensive design of signal processing including CSI sanitization, dynamic component extraction and sample augmentation, which are built into the core machine learning model. Extensive evaluation is conducted with five users for three applications, i.e., recognizing Digits, English letters, and Chinese characters, in realistic office environment. The experiment results demonstrate that the proposed system provides at least 0.9 accuracy in various combinations of users and applications with 0.93 accuracy on average.