Zeju Xu

Zeju Xu

Auto-generated profile, based on publicly available information

Zeju Xu

School of Biomedical Engineering, Shenzhen Campus of Sun Yat-sen University, Shenzhen, China

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

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

Representative Publications

View All Publications
10 May 2024 in IEEE Internet of Things Journal
Verified
Journal Article
Closed access
XYXin Yang
Xin Yang
ZXZeju Xu
Zeju Xu
HLHaodong Liu
Haodong Liu
PSPeter Shull
Peter Shull
SRStephen Redmond
Stephen Redmond
CWChanghong Wang
Changhong Wang
Human activity recognition (HAR) enables real-time monitoring of human movement, posture, and activity level, and can provide valuable information for the health management. With the continuous advancement of Internet of Things (IoT) technology, wearable sensors and smartphones equipped with various types of sensors have become widely utilized to collect multimodal data for HAR. However, in multimodal HAR, current fusion methods fall short in capturing intermodality correlations, hampering the full exploitation of complementary information between the modalities and leading to lower recognition accuracy. We thus propose a novel multiscale cross-modal interactive fusion network (MCIFN), which can fully capture correlations between various modalities and obtain an effective fused representation for HAR. Specifically, we employ a multiscale parallel convolution module to extract features from each modality at multiple scales. Then, an interactive fusion strategy based on the cross-modal attention mechanism is introduced to adjust and enhance each modality based on its correlations with other modalities. Additionally, to resolve the information redundancy caused by the interactive fusion strategy, we utilize a hybrid attention module to focus on important information in the fusion representation. Extensive experiments conducted on three publicly available data sets and one private data set demonstrate that our proposed network outperforms the previous baseline networks for HAR. Additionally, our proposed fusion strategy yielded a notable improvement in accuracy ranging from 1.87% to 9.96% compared to existing strategies. These findings imply that our newly proposed network can realize comprehensive multimodal fusion and effectively enhance HAR accuracy, potentially contributing to advancements in individual health management and personalized healthcare interventions.

14 March 2024 in IEEE Transactions on Biomedical Engineering
Verified
Journal Article
Closed access
TCTian Chen
Tian Chen
JZJingtao Zhang
Jingtao Zhang
ZXZeju Xu
Zeju Xu
SRStephen Redmond
Stephen Redmond
NLNigel Lovell
Nigel Lovell
CWChanghong Wang
Changhong Wang
Objective: Sleep apnea syndrome (SAS) is a common sleep disorder, which has been shown to be an important contributor to major neurocognitive and cardiovascular sequelae. Considering current diagnostic strategies are limited with bulky medical devices and high examination expenses, a large number of cases go undiagnosed. To enable large-scale screening for SAS, wearable photoplethysmography (PPG) technologies have been used as an early detection tool. However, existing algorithms are energy-intensive and require large amounts of memory resources, which are believed to be the major drawbacks for further promotion of wearable devices for SAS detection. Methods: In this paper, an energy-efficient method of SAS detection based on hyperdimensional computing (HDC) is proposed. Inspired by the phenomenon of chunking in cognitive psychology as a memory mechanism for improving working memory efficiency, we proposed a one-dimensional block local binary pattern (1D-BlockLBP) encoding scheme combined with HDC to preserve dominant dynamical and temporal characteristics of pulse rate signals from wearable PPG devices. Results: Our method achieved 70.17 $\%$ accuracy in sleep apnea segment detection, which is comparable with traditional machine learning methods. Additionally, our method achieves up to 67× lower memory footprint, 68× latency reduction, and 93× energy saving on the ARM Cortex-M4 processor. Conclusion: The simplicity of hypervector operations in HDC and the novel 1D-BlockLBP encoding effectively preserve pulse rate signal characteristics with high computational efficiency. Significance: This work provides a scalable solution for long-term home-based monitoring of sleep apnea, enhancing the feasibility of consistent patient care.

30 September 2024 in IEEE Transactions on Affective Computing
Verified
Journal Article
Closed access
ZXZeju Xu
Zeju Xu
GLGuanzheng Liu
Guanzheng Liu
GZGuozhen Zhao
Guozhen Zhao
ZZZhiguo Zhang
Zhiguo Zhang
CLChenzhong Li
Chenzhong Li
CWChanghong Wang
Changhong Wang
Prediction of daily mental wellbeing holds profound implications for individual healthcare and societal stability. Previous studies have shown the potential of using individual's multimodal behavioral data collected through mobile devices to predict his/her daily mental wellbeing metrics, such as stress, mood, and anxiety. However, effectively capturing long-range dependencies in behavioral time series data while accurately representing the statistical distribution patterns of various behaviors over a certain period is a significant challenge. In this paper, we propose a daily mental wellbeing prediction model based on a Topic-Guided Self-Attention Network (TGSAN). This model utilizes self-attention mechanism to capture long-range dependencies from the behavioral data collected by mobile devices. We utilize a multi-granularity time encoding method to inject time information of different granularities (i.e., day and hour, or week and day) into the behavioral data, thereby enhancing the sensibility of the self-attention network to capture every individual's habitual cyclicality rhythm. Then, we introduce a neural topic model to analyze the statistical distribution characteristics of various behaviors in the monitoring period as behavioral distribution patterns for different individuals, and further propose a topic attention network to enhance the model's classification performance by guiding the weights of long-range dependencies features from the self-attention network with the derived topic information. Compared to state-of-the-art methods, the proposed TGSAN achieved superior performance on datasets that measure different mental health indicators (stress, mood, and anxiety), with F1 scores outperforming by 4.5% and 2.3% on the Crosscheck and StudentLife datasets, respectively, and accuracy outperforming by 3.3% on the GLOBEM dataset. Our study demonstrates the effectiveness and interpretability of combining self-attention mechanisms with neural topic model, for a better understanding of the relationship between different individuals' behaviors and their mental wellbeing.

13 March 2026 in Pattern Recognition
Verified
Journal Article
Closed access
YLYi Liu
Yi Liu
ZXZeju Xu
Zeju Xu
ZLZhijun Li
Zhijun Li
ZLZhen Liang
Zhen Liang
ZZZhiguo Zhang
Zhiguo Zhang
CWChanghong Wang
Changhong Wang