Xi Zhang

Xi Zhang

Auto-generated profile, based on publicly available information

Xi Zhang

Macquarie University, Sydney, Australia

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
2 October 2023
Verified
Conference Paper
Open access
XZXi Zhang
Xi Zhang
Yu ZhangYu Zhang
Yu Zhang
Zhenguo ShiZhenguo Shi
Zhenguo Shi
TGTao Gu
Tao Gu
Facial expression recognition plays a vital role to enable emotional awareness in multimedia Internet of Things applications. Traditional camera or wearable sensor based approaches may compromise user privacy or cause discomfort. Recent device-free approaches open a promising direction by exploring Wi-Fi or ultrasound signals reflected from facial muscle movements, but limitations exist such as poor performance in presence of body motions and not being able to detect multiple targets. To bridge the gap, we propose mmFER, a novel millimeter wave (mmWave) radar based system that extracts facial muscle movements associated with mmWave signals to recognize facial expressions. We propose a novel dual-locating approach based on MIMO that explores spatial information from raw mmWave signals for face localization in space, eliminating ambient noise. In addition, collecting mmWave training data can be very costly in practice, and insufficient training dataset may lead to low accuracy. To overcome, we design a cross-domain transfer pipeline to enable effective and safe model knowledge transformation from image to mmWave. Extensive evaluations demonstrate that mmFER achieves an accuracy of 80.57% on average within a detection range between 0.3m and 2.5m, and it is robust to various real-world settings.

13 August 2021 in IEEE/ACM Transactions on Networking
Verified
Journal Article
Closed access
Yu ZhangYu Zhang
Yu Zhang
TGTao Gu
Tao Gu
XZXi Zhang
Xi Zhang
Personal mobile sensing is fast permeating our daily lives to enable activity monitoring, healthcare and rehabilitation. Combined with deep learning, these applications have achieved significant success in recent years. Different from conventional cloud-based paradigms, running deep learning on devices offers several advantages including data privacy preservation and low-latency response for both model inference and update. Since data collection is costly in reality, Google’s Federated Learning offers not only complete data privacy but also better model robustness based on data from multiple users. However, personal mobile sensing applications are mostly user-specific and highly affected by environment. As a result, continuous local changes may seriously affect the performance of a global model generated by Federated Learning. In addition, deploying Federated Learning on a local server, e.g., edge server, may quickly reach the bottleneck due to resource limitation. Towards pushing deep learning on devices, we present MDLdroid, a novel decentralized mobile deep learning framework to enable resource-aware on-device collaborative learning for personal mobile sensing applications. To address resource limitation, we propose a ChainSGD-reduce approach which includes a novel chain-directed Synchronous Stochastic Gradient Descent algorithm to effectively reduce overhead among multiple devices. We also design an agent-based multi-goal reinforcement learning mechanism to balance resources in a fair and efficient manner. Our evaluations show that our model training on off-the-shelf mobile devices achieves 2x to 3.5x faster than single-device training, and 1.5x faster on average than the existing master-slave approach.

26 February 2021 in IEEE Transactions on Mobile Computing
Verified
Journal Article
Closed access
Yu ZhangYu Zhang
Yu Zhang
TGTao Gu
Tao Gu
XZXi Zhang
Xi Zhang
Mobile deep learning (MDL) has emerged as a privacy-preserving learning paradigm for mobile devices. This paradigm offers unique features such as privacy preservation, continual learning and low-latency inference to the building of personal mobile sensing applications. However, squeezing Deep Learning to mobile devices is extremely challenging due to resource constraint. Traditional Deep Neural Networks (DNNs) are usually over-parametered, hence incurring huge resource overhead for on-device learning. In this paper, we present a novel on-device deep learning framework named MDLdroidLite that transforms traditional DNNs into resource-efficient model structures for on-device learning. To minimize resource overhead, we propose a novel release-and-inhibit control (RIC) approach based on Model Predictive Control theory to efficiently grow DNNs from tiny to backbone. We also design a gate-based fast adaptation mechanism for channel-level knowledge transformation to quickly adapt new-born neurons with existing neurons, enabling safe parameter adaptation and fast convergence for on-device training. Our evaluations show that MDLdroidLite boosts on-device training on various PMS datasets with 28× to 50× less model parameters, 4× to 10× less floating number operations than the state-of-the-art model structures while keeping the same accuracy level.

31 July 2025 in IEEE Transactions on Mobile Computing
Verified
Journal Article
Closed access
Zhenguo ShiZhenguo Shi
Zhenguo Shi
TGTao Gu
Tao Gu
Yu ZhangYu Zhang
Yu Zhang
XZXi Zhang
Xi Zhang
Blood pressure (BP) measurement is an indispensable tool in diagnosing and treating many diseases such as cardiovascular failure and stroke. Traditional direct measurement can be invasive, and wearable-based methods may have limitations of discomfort and inconvenience. Contact-free BP measurement has been recently advocated as a promising alternative. In particular, Millimeter-wave (mmWave) sensing has demonstrated its promising potential, however it is confronted with several challenges including noise and vulnerability to human’s tiny motions which may occur intentionally and inevitably. In this paper, we propose mmBP+, a contact-free mmWave-based BP measurement system with high accuracy and motion robustness. Due to the high frequency and short wavelength, mmWave signals received in the time domain are dramatically susceptible to ambient noise, and deteriorating signal quality. To reduce noise,we propose a novel approach to exploit mmWave signal’s characteristics and features in the delay-Doppler-fractional Fourier domain to significantly improve signal quality for pulse waveform construction. We also propose a periodic signal feature based functional link adaptive filter leveraging on the periodic and correlation characteristics of pulse waveform signals to alleviate the impact of human’s tiny motions. Extensive experiment results achieved by the leave-one-out cross-validation (LOOCV) method demonstrate that mmBP+ achieves the mean errors of 0.65 mmHg and 1.31 mmHg for systolic blood pressure (SBP) and diastolic blood pressure (DBP), respectively; and the standard deviation errors of 3.92 mmHg and 3.99 mmHg for SBP and DBP, respectively.

3 September 2025 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Verified
Journal Article
Open access
LWLei Wang
Lei Wang
XWXingwei Wang
Xingwei Wang
XZXi Zhang
Xi Zhang
XMXiaolei Ma
Xiaolei Ma
Yu ZhangYu Zhang
Yu Zhang
Haipeng DaiHaipeng Dai
Haipeng Dai
Speech enhancement can greatly improve the user experience during phone calls in low signal-to-noise ratio (SNR) scenarios. In this paper, we propose a low-cost, energy-efficient, and environment-independent speech enhancement system, namely AccCall, that improves phone call quality using the smartphone's built-in accelerometer. However, a significant gap remains between the underlying insight and its practical applications, as several critical challenges should be addressed, including efficiency of speech enhancement in cross-user scenario, adaptive system triggering to reduce energy consumption, and lightweight deployment for real-time processing. To this end, we first design Acc-Aided Network (AccNet), a cross-modal deep learning model inherently capable of cross-user generalization through three key components, including cross-modal fusion module, accelerometer-aided (abbreviated as acc-aided) mask generator, the unified loss function. Second, we adopt a machine learning-based approach instead of deep learning to achieve high accuracy in distinguishing call activity states followed by adaptive system triggering, ensuring lower energy consumption and efficient deployment on mobile platforms. Finally, we propose a knowledge-distillation-driven structured pruning framework that optimizes model efficiency while preserving performance. Extensive experiments with 20 participants have been conducted under a user-independent scenario. The results show that AccCall achieves excellent and reliable adaptive triggering performance, and enables substantial real-time improvements in SISDR, SISNR, STOI, PESQ, and WER, demonstrating the superiority of our system in enhancing speech quality and intelligibility for phone calls.

30 September 2026 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Verified
Journal Article
Open access
LWLei Wang
Lei Wang
JSJiangxuan Shen
Jiangxuan Shen
XZXi Zhang
Xi Zhang
DZDalin Zhang
Dalin Zhang
JLJingyu Li
Jingyu Li
HHe Huang
He Huang
The widespread use of earphones has enabled various sensing applications, including activity recognition, health monitoring, and context-aware computing. Among these, earphone-based user authentication has become a key technique by leveraging unique biometric features. However, existing earphone-based authentication systems face key limitations: they either require explicit user interaction or active speaker output, or suffer from poor accessibility and vulnerability to environmental noise, which hinders large-scale deployment. In this paper, we propose a passive authentication system, called AccLock, which leverages distinctive features extracted from in-ear BCG signals to enable secure and unobtrusive user verification. Our system offers several advantages over previous systems, including zero-involvement for both the device and the user, ubiquitous, and resilient to environmental noise. To realize this, we first design a two-stage denoising scheme to suppress both inherent and sporadic interference. To extract user-specific features, we then propose a disentanglement-based deep learning model, HIDNet, which explicitly separates user-specific features from shared nuisance components. Lastly, we develop a scalable authentication framework based on a Siamese network that eliminates the need for per-user classifier training. We conduct extensive experiments with 33 participants, achieving an average FAR of 3.13% and FRR of 2.99%, which demonstrates the practical feasibility of AccLock.