Xiaolin Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Xiaolin Li has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Deepu John
University College Dublin, Dublin, Ireland
4 shared publications - Barry Cardiff
University College Dublin
3 shared publications - Guoxin Wang
Trinity College Dublin
2 shared publications - Adnan Ashraf
ZiO Health Ltd
1 shared publication - Binhua Huang
Guangdong Provincial Key Laboratory of Robotics and Intelligent System, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China
1 shared publication Dawei Mai
Microelectronic Circuits Centre Ireland, University College Dublin,Belfield, Dublin 4,Ireland
1 shared publication
Latest Publications
- Open Access
- 9 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Translational Engineering in Health and Medicine
- Vol. 14, 285-297
Neural Architecture Search-Based ECG Classification for Low-Power Edge Deployment in IoT Healthcare Systems
Objective: To develop and validate a deployment-centric ECG arrhythmia classification framework for low-power wearable edge devices, with emphasis on efficient on-device screening and reduced communication burden in remote cardiac monitoring workflows, optimized for seamless integration into clinical remote monitoring workflows using resource-constrained wearable IoT edge devices. Methods and procedures: We design a clinically-oriented wearable-to-phone workflow in which the edge device performs on-device ECG inference and transmits actionable anomaly summaries to a smartphone via Bluetooth Low Energy (BLE). This approach reduces latency, energy cost, and privacy risks compared to continuous raw-signal transmission. A NAS process tailored to arrhythmia detection automatically explores candidate architectures under diverse compute and memory constraints to identify feasible models for embedded platforms. Multiple deployment-ready variants are generated via quantization and pruning, and implemented on an ARM Cortex-M33-based system. System efficiency is quantified through end-to-end measurements of energy consumption and data throughput, ensuring its readiness for long-term ambulatory monitoring. Results: The deployed models achieve real-time inference on an ARM Cortex-M33 with as little as 21.7 kB RAM and 265k MACs. After optimization, energy consumption is reduced to 5.51 mJ per inference. Compared with the original model, RAM usage is reduced by 53.6%, with only a 0.5% absolute decrease in accuracy (97.9% to 97.4%). Hardware-in-the-loop power profiling confirms significant efficiency gains, validating the clinical feasibility of the proposed selective offloading strategy. Conclusion: A NAS-guided, deployment-centric pipeline-spanning architecture search, compression/mapping, and embedded power profiling enables accurate, real-time ECG classification within stringent wearable constraints. The selective-offloading strategy bridges the gap between laboratory algorithms and practical cardiac surveillance by providing a high-fidelity, low-power, and clinically-aware solution for continuous arrhythmia screening. Clinical impact: This work supports the design of low-power wearable arrhythmia screening systems that can perform beat-level on-device inference and transmit compact abnormal-event summaries to a companion smartphone for downstream review. By reducing unnecessary wireless transmissions while maintaining timely anomaly reporting, the proposed framework may support battery-aware, privacy-enhancing remote cardiac monitoring in home and outpatient settings. The system is intended as a screening and triage aid rather than a standalone diagnostic tool. Clinical and Translational Impact Statement: An energy-efficient on-wearable ECG arrhythmia classifier sends BLE anomaly summaries to a phone, enabling privacy-preserving home monitoring and scalable integration into remote cardiac surveillance workflows.
- 9 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Translational Engineering in Health and Medicine
- Vol. 14, 285-297
Open AccessNeural Architecture Search-Based ECG Classification for Low-Power Edge Deployment in IoT Healthcare Systems
Objective: To develop and validate a deployment-centric ECG arrhythmia classification framework for low-power wearable edge devices, with emphasis on efficient on-device screening and reduced communication burden in remote cardiac monitoring workflows, optimized for seamless integration into clinical remote monitoring workflows using resource-constrained wearable IoT edge devices. Methods and procedures: We design a clinically-oriented wearable-to-phone workflow in which the edge device performs on-device ECG inference and transmits actionable anomaly summaries to a smartphone via Bluetooth Low Energy (BLE). This approach reduces latency, energy cost, and privacy risks compared to continuous raw-signal transmission. A NAS process tailored to arrhythmia detection automatically explores candidate architectures under diverse compute and memory constraints to identify feasible models for embedded platforms. Multiple deployment-ready variants are generated via quantization and pruning, and implemented on an ARM Cortex-M33-based system. System efficiency is quantified through end-to-end measurements of energy consumption and data throughput, ensuring its readiness for long-term ambulatory monitoring. Results: The deployed models achieve real-time inference on an ARM Cortex-M33 with as little as 21.7 kB RAM and 265k MACs. After optimization, energy consumption is reduced to 5.51 mJ per inference. Compared with the original model, RAM usage is reduced by 53.6%, with only a 0.5% absolute decrease in accuracy (97.9% to 97.4%). Hardware-in-the-loop power profiling confirms significant efficiency gains, validating the clinical feasibility of the proposed selective offloading strategy. Conclusion: A NAS-guided, deployment-centric pipeline-spanning architecture search, compression/mapping, and embedded power profiling enables accurate, real-time ECG classification within stringent wearable constraints. The selective-offloading strategy bridges the gap between laboratory algorithms and practical cardiac surveillance by providing a high-fidelity, low-power, and clinically-aware solution for continuous arrhythmia screening. Clinical impact: This work supports the design of low-power wearable arrhythmia screening systems that can perform beat-level on-device inference and transmit compact abnormal-event summaries to a companion smartphone for downstream review. By reducing unnecessary wireless transmissions while maintaining timely anomaly reporting, the proposed framework may support battery-aware, privacy-enhancing remote cardiac monitoring in home and outpatient settings. The system is intended as a screening and triage aid rather than a standalone diagnostic tool. Clinical and Translational Impact Statement: An energy-efficient on-wearable ECG arrhythmia classifier sends BLE anomaly summaries to a phone, enabling privacy-preserving home monitoring and scalable integration into remote cardiac surveillance workflows.
- 16 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (11) , 24938-24953
Multiagent Collaborative Inference Optimization for Large-Scale DNNs in IoT Edge Systems
Deploying large deep neural networks (DNNs) on resource-constrained Internet of Things (IoT) devices is often limited by high inference latency and energy consumption. This article studies collaborative inference between user devices and edge servers (ESs), where a deep neural network (DNN) is partitioned and executed across both sides under shared wireless resources. We jointly optimize the partition point across layers, uplink channel allocation and transmission power control, and formulate the coupled decisions as a Markov decision process (MDP). Under centralized training with decentralized execution (CTDE), we develop a multiagent proximal policy optimization (MAPPO) framework with a multihead critic and profiling-aware state design to stabilize learning under large hybrid action spaces. To validate practicality, we implement a containerized experimental platform that emulates resource-limited IoT devices and GPU-enabled ESs, enabling repeatable measurements of end-to-end latency and energy. Experiments with representative DNNs show that the proposed method consistently improves the latency–energy tradeoff over baseline algorithms, achieving lower end-to-end latency and comparable or lower energy consumption across diverse model structures and computation patterns.
- 16 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (11) , 24938-24953
Multiagent Collaborative Inference Optimization for Large-Scale DNNs in IoT Edge Systems
Deploying large deep neural networks (DNNs) on resource-constrained Internet of Things (IoT) devices is often limited by high inference latency and energy consumption. This article studies collaborative inference between user devices and edge servers (ESs), where a deep neural network (DNN) is partitioned and executed across both sides under shared wireless resources. We jointly optimize the partition point across layers, uplink channel allocation and transmission power control, and formulate the coupled decisions as a Markov decision process (MDP). Under centralized training with decentralized execution (CTDE), we develop a multiagent proximal policy optimization (MAPPO) framework with a multihead critic and profiling-aware state design to stabilize learning under large hybrid action spaces. To validate practicality, we implement a containerized experimental platform that emulates resource-limited IoT devices and GPU-enabled ESs, enabling repeatable measurements of end-to-end latency and energy. Experiments with representative DNNs show that the proposed method consistently improves the latency–energy tradeoff over baseline algorithms, achieving lower end-to-end latency and comparable or lower energy consumption across diverse model structures and computation patterns.
- 15 January 2025
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 104
DCentNet: Decentralized multistage biomedical signal classification using early exits
- 15 January 2025
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 104
DCentNet: Decentralized multistage biomedical signal classification using early exits
- Open Access
- 4 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 24 (14) , 23324-23335
A Two-Stage ECG Classifier for Decentralized Inferencing Across Edge-Cloud Continuum
In this paper, we propose a multi-stage electrocardiogram (ECG) classifier for distributed machine learning (ML) inferencing across the edge-cloud continuum for wearable systems. Traditionally, biomedical data acquired from wearable systems are processed in one step, using a single-stage classifier deployed either on a cloud or on an edge device. Though there are merits, both approaches have limitations that relate to computational complexity, network connectivity, etc. In this work, we propose a multi-stage, cascaded, ensemble classifier that aims to address these limitations by decentralizing the processing steps, while achieving good overall performance. We employed low-complexity, explainable boosting machines (EBM) and convolutional neural networks (CNN) to develop a multi-stage distributed ECG classifier, which achieves a high-sensitivity binary classification on the edge device and a more comprehensive multi-class classification on the cloud. In standalone performance evaluation using the MIT-BIH Arrhythmia database, the Stage-1 EBM classifier and Stage-2 CNN classifier achieved a maximum accuracy, sensitivity of 96.71%, 96.76%, and 99.49%, 98.19% respectively. Further, the distributed multi-stage classifier achieved a maximum cumulative binary classification accuracy, sensitivity of 99.64%, 99.01%, and multi-class classification accuracy, sensitivity of 99.56%, 98.79% when DtC equals 40%. Further, we evaluated the use of the EBM classifier threshold as a control parameter to dynamically vary the system performance and network traffic based on real-time conditions. We verified the feasibility of the model, calculated the energy consumption, and estimated the latency. When streaming only 40% of the data to the cloud, it will result in 60% latency saving. With the proposed technique, energy consumption is reduced by approximately 3 times.
- 4 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 24 (14) , 23324-23335
Open AccessA Two-Stage ECG Classifier for Decentralized Inferencing Across Edge-Cloud Continuum
In this paper, we propose a multi-stage electrocardiogram (ECG) classifier for distributed machine learning (ML) inferencing across the edge-cloud continuum for wearable systems. Traditionally, biomedical data acquired from wearable systems are processed in one step, using a single-stage classifier deployed either on a cloud or on an edge device. Though there are merits, both approaches have limitations that relate to computational complexity, network connectivity, etc. In this work, we propose a multi-stage, cascaded, ensemble classifier that aims to address these limitations by decentralizing the processing steps, while achieving good overall performance. We employed low-complexity, explainable boosting machines (EBM) and convolutional neural networks (CNN) to develop a multi-stage distributed ECG classifier, which achieves a high-sensitivity binary classification on the edge device and a more comprehensive multi-class classification on the cloud. In standalone performance evaluation using the MIT-BIH Arrhythmia database, the Stage-1 EBM classifier and Stage-2 CNN classifier achieved a maximum accuracy, sensitivity of 96.71%, 96.76%, and 99.49%, 98.19% respectively. Further, the distributed multi-stage classifier achieved a maximum cumulative binary classification accuracy, sensitivity of 99.64%, 99.01%, and multi-class classification accuracy, sensitivity of 99.56%, 98.79% when DtC equals 40%. Further, we evaluated the use of the EBM classifier threshold as a control parameter to dynamically vary the system performance and network traffic based on real-time conditions. We verified the feasibility of the model, calculated the energy consumption, and estimated the latency. When streaming only 40% of the data to the cloud, it will result in 60% latency saving. With the proposed technique, energy consumption is reduced by approximately 3 times.
- 1 June 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Experimental Confirmation of Wandering Spurs in a Commercial Fractional-N Frequency Synthesizer with a MASH 1-1-1 Divider Controller
Fractional-N frequency synthesizers can exhibit a phenomenon called wandering spurs which is undesirable in many applications. This paper presents experimental confirmation of the behavior in a fractional- $N$ frequency synthesizer with a MASH 1-1-1-based divider controller.- 1 June 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Experimental Confirmation of Wandering Spurs in a Commercial Fractional-N Frequency Synthesizer with a MASH 1-1-1 Divider Controller
Fractional-N frequency synthesizers can exhibit a phenomenon called wandering spurs which is undesirable in many applications. This paper presents experimental confirmation of the behavior in a fractional- $N$ frequency synthesizer with a MASH 1-1-1-based divider controller.- 23 April 2019
- Published by SPIE-Intl Soc Optical Eng
- Vol. 11030, 48
Convolution theorems for the linear canonical transforms
The linear canonical transforms (LCTs) are a Lie group of transforms including the Fresnel and Fourier transforms that describe scalar wave propagation in quadratic phase systems. As such, they are useful in system analysis and design, and their discretisations are important for opto-numerical systems, e.g. numerical reconstruction algorithms in digital holography. An important topic in the literature is therefore the generalization of Fourier transform properties for the LCTs. A number of authors have proposed convolution theorems for the linear canonical transform, with different goals in mind. In this paper, we compare those methods, with particular attention being paid to the consequences of discretization. In a similar way to how discrete convolution associated with the DFT differs from that associated with the Fourier transform, we must take the chirp-periodic nature of discrete LCTs into account when determining the discrete convolution associated with LCTs. This work is of significance for the simulation of VanderLugt correlators, which have been used for optical implementations of neural networks, and for optical filtering operations and coherent optical signal processing in general.
- 23 April 2019
- Published by SPIE-Intl Soc Optical Eng
- Vol. 11030, 48
Convolution theorems for the linear canonical transforms
The linear canonical transforms (LCTs) are a Lie group of transforms including the Fresnel and Fourier transforms that describe scalar wave propagation in quadratic phase systems. As such, they are useful in system analysis and design, and their discretisations are important for opto-numerical systems, e.g. numerical reconstruction algorithms in digital holography. An important topic in the literature is therefore the generalization of Fourier transform properties for the LCTs. A number of authors have proposed convolution theorems for the linear canonical transform, with different goals in mind. In this paper, we compare those methods, with particular attention being paid to the consequences of discretization. In a similar way to how discrete convolution associated with the DFT differs from that associated with the Fourier transform, we must take the chirp-periodic nature of discrete LCTs into account when determining the discrete convolution associated with LCTs. This work is of significance for the simulation of VanderLugt correlators, which have been used for optical implementations of neural networks, and for optical filtering operations and coherent optical signal processing in general.
- 1 November 2013
- Published by China Science Publishing & Media Ltd. in Chinese Journal of Applied Chemistry
- Vol. 30 (11) , 1237
Applications of Ugi Reaction to the Synthesis of Natural Products
- 1 November 2013
- Published by China Science Publishing & Media Ltd. in Chinese Journal of Applied Chemistry
- Vol. 30 (11) , 1237
Applications of Ugi Reaction to the Synthesis of Natural Products