Xiangmin Luo

Xiangmin Luo

Auto-generated profile, based on publicly available information

Xiangmin Luo

School of Electronic and Information Engineering, Guangdong Polytechnic Normal University, Guangzhou, China

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

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

Representative Publications

View All Publications
2 October 2023 in IEEE Sensors Journal
Verified
Journal Article
Closed access
FLFangyuan LEI
Fangyuan LEI
XLXiangmin Luo
Xiangmin Luo
ZCZiwei Chen
Ziwei Chen
HZHoumian Zhou
Houmian Zhou
Graph neural networks demonstrate effectiveness in fault diagnosis due to their capability to handle complex dependencies and nonlinear mappings. However, current learning methods relying on simple graphs fail to fully exploit the representation power between nodes when encountering fault data in time series. Moreover, the information extraction capacity of high-order neighbors is limited. Existing approaches exhibit suboptimal performance when confronted with noise interference and a scarcity of training data. To tackle these challenges, we propose a signal feature extraction method based on a dual-channel wavelet convolutional network and hypergraph convolutional network (WC-HGCN) for fault type classification. In WC-HGCN, fault data is represented as a hypergraph, enabling the learning of intricate high-order relationships between data through hypergraph convolution. We use wavelet convolution to extract sample trend and noise information, which can effectively solve the problem of noise interference and small samples, and increase the robustness of the model. Finally, we evaluate WC-HGCN on publicly available fault diagnosis datasets, and the results demonstrate its state-of-the-art performance.

20 May 2024 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
XLXiangmin Luo
Xiangmin Luo
ZCZiwei Chen
Ziwei Chen
DHDa Huang
Da Huang
FLFangyuan LEI
Fangyuan LEI
CWChang-Dong Wang
Chang-Dong Wang
ILIman Yi Liao
Iman Yi Liao
Sensors ensure the normal operation of the system by monitoring and collecting large amounts of data in real time. Due to the complexity and variability of the working environment, the industrial system exhibits different behaviors under different working conditions. Existing research methods mainly rely on graph structure to learn one-to-one connections between samples, and there is a lack of research on higher-order features and contextual data. And changes in the current moment of time series data may have the same impact on the next moment and some time in the future. Therefore, we propose a multi-channel variational hypergraph auto-encoder network(MC-VHAE) for unsupervised domain adaptation fault diagnosis across working conditions. In the proposed multi-channel network, on the one hand, convolutional neural networks are used to extract multi-scale features from time-domain fault data. On the other hand, discrete wavelet transform is used to separate the high-frequency and low-frequency components of the fault sample in the time-frequency domain. The potential higher order information is extracted by variational hypergraph auto-encoder (VHAE). In VHAE, a multi-order neighborhood hypergraph convolutional layer (MON-HGCL) is designed to aggregate the high-order feature information of different order neighborhoods in the hypergraph nodes. Finally, the feature fusion layer is used to retain the low-frequency trend component and multi-scale features while removing high-frequency noise components. Experimental results show that MC-VHAE outperforms existing methods and demonstrates its ability to extract domain-invariant features under different operating conditions.

9 June 2025 in Neurocomputing
Verified
Journal Article
Closed access
ZCZiwei Chen
Ziwei Chen
JJJianjian Jiang
Jianjian Jiang
XLXiangmin Luo
Xiangmin Luo
FLFangyuan Lei
Fangyuan Lei
XYXiaochen Yuan
Xiaochen Yuan
JZJin Zhan
Jin Zhan

9 March 2026 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
JGJie Gao
Jie Gao
ISIeong Weng San
Ieong Weng San
XLXiangmin Luo
Xiangmin Luo
ZCZhengxuan Chen
Zhengxuan Chen
ZYZitong Yu
Zitong Yu
YSYue Sun
Yue Sun
Remote photoplethysmography (rPPG) offers a non-invasive solution for estimating critical vital signs, such as heart rate (HR) and respiratory rate (RR). To improve algorithmic generalizability, domain generalization has garnered increasing attention for rPPG. However, existing approaches often overlook the frequency periodicity driven by blood-volume pulsations and the dependencies among tasks, which limits generalization. To address this, PhysVLM is proposed as a cross-domain, multi-task rPPG framework. First, a pre-trained vision language model provides frequency-related priors. These priors are injected into a visual-only student network via knowledge distillation. This encourages the student to focus on physiologically driven periodic features. Next, the wavelet temporal block (WTB) is introduced for multi-scale time-frequency decomposition and adaptive filtering to suppress task irrelevant noise. In addition, frequency-aligned spatial attention (FASA) is proposed to perform frequency-aligned bandpass selection within rhythmically prominent regions. Cross-task consistency constraints are incorporated to reduce spectral leakage. Extensive experiments on five public datasets, together with clinical validation on 25 preterm infants in the neonatal intensive care unit, show that PhysVLM achieves stable improvements in cross-domain evaluations. On the challenging NBHR dataset, the proposed method reduces HR MAE from 5.85 bpm to 2.85 bpm. This framework provides a robust foundation for reliable rPPG applications in clinical settings.