Qinglin Zhao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Qinglin Zhao has a total of 67 co-authors. Here are the top co-authors by number of shared publications:
- Bin Hu
Beijing Institute of Technology
16 shared publications - Fuze Tian
School of Information Science and Engineering, Lanzhou University, Lanzhou, China
14 shared publications - Lixin Zhang
School of Information Science and Engineering, Lanzhou University, Lanzhou, China
10 shared publications - Mingqi Zhao
School of Information Science and Engineering, Lanzhou University, Lanzhou, China
8 shared publications - Hua Jiang
School of Information Science and Engineering, Lanzhou University, Lanzhou, China
6 shared publications - Kunbo Cui
School of Information Science and Engineering, Lanzhou University, Lanzhou, China
6 shared publications
Latest Publications
- 22 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cognitive and Developmental Systems
- Vol. PP (23798920) , 1-17
An EEG Data-Driven Neural Mass Model for Investigating Abnormal Reward Circuitry of Heroin Addicts
Heroin abuse induces plastic and adaptive changes in the brain’s reward circuitry, leading to compulsive drugseeking behavior. While neuroimaging techniques have revealed some abnormal physiological parameters, measuring key neural interactions remains challenging. This study employs a datadriven Neural Mass Model (NMM) to characterize neural dynamics within the nucleus accumbens (NAc) and identify potential biomarkers of heroin addiction. The model simulates interactions among neurons in the ventral tegmental area (VTA), nucleus accumbens (NAc), and medial prefrontal cortex (mPFC), with parameters optimized using particle swarm optimization (PSO). NAc source signals were reconstructed from electroencephalography (EEG) recordings using a 12-layer realistic head model. Model validation confirmed robust performance in capturing NAc dynamics within low-frequency bands (δ, θ, α) through time-domain, frequency-domain, and nonlinear feature analyses. Systematic parameter optimization identified four candidate biomarkers distinguishing heroin addicts from healthy controls: bidirectional synaptic connectivity between VTA and NAc (C3, C4), maximum firing rate of NAc neurons (eNAc), and average synaptic gain (HNAc). These parameters exhibited significant correlations with clinical variables including abstinence duration and heroin dosage. Mediation, sensitivity-specificity, and constraint analyses confirmed that the identified biomarkers reflect genuine neurophysiological alterations rather than artifacts of model simplification. These computational findings converge with electrophysiological, optogenetic, and pharmacological evidence, supporting the model’s neurobiological validity. This study provides a mesoscopic computational framework for understanding addiction-related neuroplasticity and informs future development of targeted neuromodulation interventions.
- 22 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cognitive and Developmental Systems
- Vol. PP (23798920) , 1-17
An EEG Data-Driven Neural Mass Model for Investigating Abnormal Reward Circuitry of Heroin Addicts
Heroin abuse induces plastic and adaptive changes in the brain’s reward circuitry, leading to compulsive drugseeking behavior. While neuroimaging techniques have revealed some abnormal physiological parameters, measuring key neural interactions remains challenging. This study employs a datadriven Neural Mass Model (NMM) to characterize neural dynamics within the nucleus accumbens (NAc) and identify potential biomarkers of heroin addiction. The model simulates interactions among neurons in the ventral tegmental area (VTA), nucleus accumbens (NAc), and medial prefrontal cortex (mPFC), with parameters optimized using particle swarm optimization (PSO). NAc source signals were reconstructed from electroencephalography (EEG) recordings using a 12-layer realistic head model. Model validation confirmed robust performance in capturing NAc dynamics within low-frequency bands (δ, θ, α) through time-domain, frequency-domain, and nonlinear feature analyses. Systematic parameter optimization identified four candidate biomarkers distinguishing heroin addicts from healthy controls: bidirectional synaptic connectivity between VTA and NAc (C3, C4), maximum firing rate of NAc neurons (eNAc), and average synaptic gain (HNAc). These parameters exhibited significant correlations with clinical variables including abstinence duration and heroin dosage. Mediation, sensitivity-specificity, and constraint analyses confirmed that the identified biomarkers reflect genuine neurophysiological alterations rather than artifacts of model simplification. These computational findings converge with electrophysiological, optogenetic, and pharmacological evidence, supporting the model’s neurobiological validity. This study provides a mesoscopic computational framework for understanding addiction-related neuroplasticity and informs future development of targeted neuromodulation interventions.
- 6 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. 13 (3) , 2899-2911
Dynamic Evolution of Prefrontal Neural Activity in Depression: A Bayesian Probability Model
Depression is a common emotional disorder in modern society that causes growing burdens globally. This mental disorder has been frequently confirmed to be closely related to abnormalities in the prefrontal cortex (PFC). However, it remains to be fully investigated whether the neural activity in the PFC has regular quasi-steady spatiotemporal structures, and whether dynamic patterns of these structures are associated with prefrontal dysfunction in depression. To further uncover such neural correlates, we extended the traditional electroencephalography (EEG) microstates to a novel localized level and developed a variational Bayesian probabilistic generative model to decode such localized microstate patterns from few-channel prefrontal EEG signals. We validated the method with a publicly available multichannel EEG dataset obtained from 165 healthy individuals and a three-channel prefrontal EEG dataset (43 depressed and 43 healthy). The approach was finally used to examine dynamic evolution of prefrontal neural activities in depression. Our results demonstrate that the localized prefrontal microstates exhibit high cross-dataset reproducibility and were characterized by finer spatiotemporal patterns independent from traditional whole-brain microstates. The results further revealed significant emotional task-specific abnormalities in localized prefrontal microstates between the depressed and the healthy individuals, including more frequent occurrences and shorter durations in high-power bilaterally asymmetric microstates, as well as less organized low-power symmetric microstates. Our extended concept of the localized microstates and associated analytical methods provides a novel theoretical framework for elucidating prefrontal neural dynamics and also lays a theoretical foundation for uncovering prefrontal functional abnormalities in depression and developing auxiliary diagnostic tools with prefrontal few-channel EEG data.
- 6 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. 13 (3) , 2899-2911
Dynamic Evolution of Prefrontal Neural Activity in Depression: A Bayesian Probability Model
Depression is a common emotional disorder in modern society that causes growing burdens globally. This mental disorder has been frequently confirmed to be closely related to abnormalities in the prefrontal cortex (PFC). However, it remains to be fully investigated whether the neural activity in the PFC has regular quasi-steady spatiotemporal structures, and whether dynamic patterns of these structures are associated with prefrontal dysfunction in depression. To further uncover such neural correlates, we extended the traditional electroencephalography (EEG) microstates to a novel localized level and developed a variational Bayesian probabilistic generative model to decode such localized microstate patterns from few-channel prefrontal EEG signals. We validated the method with a publicly available multichannel EEG dataset obtained from 165 healthy individuals and a three-channel prefrontal EEG dataset (43 depressed and 43 healthy). The approach was finally used to examine dynamic evolution of prefrontal neural activities in depression. Our results demonstrate that the localized prefrontal microstates exhibit high cross-dataset reproducibility and were characterized by finer spatiotemporal patterns independent from traditional whole-brain microstates. The results further revealed significant emotional task-specific abnormalities in localized prefrontal microstates between the depressed and the healthy individuals, including more frequent occurrences and shorter durations in high-power bilaterally asymmetric microstates, as well as less organized low-power symmetric microstates. Our extended concept of the localized microstates and associated analytical methods provides a novel theoretical framework for elucidating prefrontal neural dynamics and also lays a theoretical foundation for uncovering prefrontal functional abnormalities in depression and developing auxiliary diagnostic tools with prefrontal few-channel EEG data.
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Systems and Rehabilitation Engineering
- Vol. 34, 4164-4178
MF-PNet: A Lightweight Multi-Frequency Progressive Neural Network for Depression Classification Deployed on Wearable EEG Sensors
Low-channel wearable EEG offers a compact acquisition configuration for depression recognition, but its limited spatial coverage and the computational constraints of edge devices present challenges for representation learning and embedded inference. This paper presents MF-PNet, a multi-frequency progressive neural network designed for three-channel EEG. Five frequency-specific branches process the canonical EEG bands in parallel, while a full-band branch progressively aggregates their representations through layer-wise lateral transfer and squeeze-and-excitation-based channel recalibration. Temporal and channel attention are further incorporated to emphasize informative EEG segments and frontal channels. Subject-level leave-one-subject-out cross-validation was conducted on a balanced in-house cohort of 146 participants. Across five random seeds, MF-PNet achieved an accuracy and F1-score of 98.8% ± 1.0% under auditory stimulation. Within-dataset evaluation on the independent Figshare and MODMA cohorts yielded accuracies of 92.1%±2.0% and 66.4%±6.5%, respectively. For embedded implementation, MF-PNet was distilled into a MobileNetV3-based student containing 75.51K parameters. The student achieved an accuracy of 99.3% ± 1.2%, and its INT8 representation retained an accuracy of 98.5% ± 0.9%. On the STM32U575CGT6 MCU, the implementation required 217.15KB of ROM and 215.78KB of RAM, with an inference latency of 45.99ms per 2-s EEG window and a power consumption of 43.2mW. The resulting framework integrates progressive cross-band EEG representation learning, teacher-student compression, INT8 quantization, and MCU execution into a unified pipeline for three-channel wearable EEG depression recognition.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Systems and Rehabilitation Engineering
- Vol. 34, 4164-4178
Open AccessMF-PNet: A Lightweight Multi-Frequency Progressive Neural Network for Depression Classification Deployed on Wearable EEG Sensors
Low-channel wearable EEG offers a compact acquisition configuration for depression recognition, but its limited spatial coverage and the computational constraints of edge devices present challenges for representation learning and embedded inference. This paper presents MF-PNet, a multi-frequency progressive neural network designed for three-channel EEG. Five frequency-specific branches process the canonical EEG bands in parallel, while a full-band branch progressively aggregates their representations through layer-wise lateral transfer and squeeze-and-excitation-based channel recalibration. Temporal and channel attention are further incorporated to emphasize informative EEG segments and frontal channels. Subject-level leave-one-subject-out cross-validation was conducted on a balanced in-house cohort of 146 participants. Across five random seeds, MF-PNet achieved an accuracy and F1-score of 98.8% ± 1.0% under auditory stimulation. Within-dataset evaluation on the independent Figshare and MODMA cohorts yielded accuracies of 92.1%±2.0% and 66.4%±6.5%, respectively. For embedded implementation, MF-PNet was distilled into a MobileNetV3-based student containing 75.51K parameters. The student achieved an accuracy of 99.3% ± 1.2%, and its INT8 representation retained an accuracy of 98.5% ± 0.9%. On the STM32U575CGT6 MCU, the implementation required 217.15KB of ROM and 215.78KB of RAM, with an inference latency of 45.99ms per 2-s EEG window and a power consumption of 43.2mW. The resulting framework integrates progressive cross-band EEG representation learning, teacher-student compression, INT8 quantization, and MCU execution into a unified pipeline for three-channel wearable EEG depression recognition.
- 15 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. 13 (2) , 2450-2460
Hyper-Parallel Superscalar Asynchronous RISC-V Processor Based on Event-Driven Logic
Event-driven neuromorphic computing involves sparse and asynchronous signal activity, which leads to irregular computation patterns and fine-grained concurrency. As a result, processing architectures need to support both high parallelism and energy efficiency. Among existing architectural solutions, superscalar designs exhibit significant potential for addressing high parallelism demands. However, conventional superscalar processors, which rely on synchronous circuits, maintain high-frequency clocking at all times, leading to substantial power inefficiency in sparse computation scenarios. To address this issue, we propose an asynchronous superscalar architecture that replaces global clocking with fully local handshake-based control, implemented using a bundled-data asynchronous protocol. The design supports decoding of up to 64 scalar instructions per cycle and implements the RISC-V RV32IMC instruction set. A prototype was fabricated using a 110 nm complementary metal oxide semiconductor (CMOS) process and was evaluated through post-layout simulation. Operating at 1.2 V, the processor delivers a peak INT8 throughput of 669.4 GOPS, with a static power consumption of 421 mW.
- 15 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. 13 (2) , 2450-2460
Hyper-Parallel Superscalar Asynchronous RISC-V Processor Based on Event-Driven Logic
Event-driven neuromorphic computing involves sparse and asynchronous signal activity, which leads to irregular computation patterns and fine-grained concurrency. As a result, processing architectures need to support both high parallelism and energy efficiency. Among existing architectural solutions, superscalar designs exhibit significant potential for addressing high parallelism demands. However, conventional superscalar processors, which rely on synchronous circuits, maintain high-frequency clocking at all times, leading to substantial power inefficiency in sparse computation scenarios. To address this issue, we propose an asynchronous superscalar architecture that replaces global clocking with fully local handshake-based control, implemented using a bundled-data asynchronous protocol. The design supports decoding of up to 64 scalar instructions per cycle and implements the RISC-V RV32IMC instruction set. A prototype was fabricated using a 110 nm complementary metal oxide semiconductor (CMOS) process and was evaluated through post-layout simulation. Operating at 1.2 V, the processor delivers a peak INT8 throughput of 669.4 GOPS, with a static power consumption of 421 mW.
- 26 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 37 (3) , 600-617
MAMILS: A Memory-Aware Multiobjective Scheduler for Real-Time Embedded EEG Depression Diagnosis
Depression detection using Electroencephalogram (EEG) signals obtained from wearable medical-assisted diagnostic systems has become a well-established approach in the field of affective disorders. However, despite recent advancements, on-board Artificial Intelligence (AI) models still demand substantial computational resources, presenting significant challenges for deployment on resource-constrained wearable medical devices. Embedded Multi-core Processors (MPs) offer a promising solution for accelerating these models. However, the limited computational capabilities of embedded MPs, combined with the structural diversity of AI models, complicate resource allocation and increase associated costs. To address these challenges, we propose a Memory-Aware Multi-Objective Iterative Local Search (MAMILS) algorithm to optimize task scheduling, thereby improving the efficiency of AI model deployment on wearable EEG devices. Experimental results across seven AI models demonstrate that, the MAMILS approach yields substantial improvements in key performance indicators: Total Energy Consumption ($\bm {TEC}$ ) with an average reduction of 47.57%,$\bm {Makespan}$ with an average reduction of 48.75%, and$\bm {Throughput}$ with an average increase of 198.37%, all while maintaining satisfactory classification performance for both Machine Learning (ML) and Deep Learning (DL) models. Especially, on-board deployment of EEGNeX achieves an accuracy of 93.4%, sensitivity of 91.6%, and specificity of 95.8%. Further analysis indicates that, when integrated with wearable EEG sensors and executable on-board AI models, the proposed MAMILS optimization strategy shows significant promise in facilitating the widespread adoption of low-power, real-time diagnostic systems for depression detection.- 26 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 37 (3) , 600-617
MAMILS: A Memory-Aware Multiobjective Scheduler for Real-Time Embedded EEG Depression Diagnosis
Depression detection using Electroencephalogram (EEG) signals obtained from wearable medical-assisted diagnostic systems has become a well-established approach in the field of affective disorders. However, despite recent advancements, on-board Artificial Intelligence (AI) models still demand substantial computational resources, presenting significant challenges for deployment on resource-constrained wearable medical devices. Embedded Multi-core Processors (MPs) offer a promising solution for accelerating these models. However, the limited computational capabilities of embedded MPs, combined with the structural diversity of AI models, complicate resource allocation and increase associated costs. To address these challenges, we propose a Memory-Aware Multi-Objective Iterative Local Search (MAMILS) algorithm to optimize task scheduling, thereby improving the efficiency of AI model deployment on wearable EEG devices. Experimental results across seven AI models demonstrate that, the MAMILS approach yields substantial improvements in key performance indicators: Total Energy Consumption ($\bm {TEC}$ ) with an average reduction of 47.57%,$\bm {Makespan}$ with an average reduction of 48.75%, and$\bm {Throughput}$ with an average increase of 198.37%, all while maintaining satisfactory classification performance for both Machine Learning (ML) and Deep Learning (DL) models. Especially, on-board deployment of EEGNeX achieves an accuracy of 93.4%, sensitivity of 91.6%, and specificity of 95.8%. Further analysis indicates that, when integrated with wearable EEG sensors and executable on-board AI models, the proposed MAMILS optimization strategy shows significant promise in facilitating the widespread adoption of low-power, real-time diagnostic systems for depression detection.- 28 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. 17 (1) , 511-528
FMEFF Mechanism: A FastDTW-Based Music-EEG Feature Fusion Approach for Identifying Enjoyment Levels in Music Therapy
Research indicates that individuals who experience greater music enjoyment exhibit stronger brain–music synchronization, which is positively correlated with improved therapeutic outcomes in depression treatment. However, most existing studies on music therapy predominantly focus on emotional states, often overlooking enjoyment levels. To address this gap, we propose a novel fusion mechanism based on Fast Dynamic Time Warping (FastDTW) to objectively assess music enjoyment. First, we introduce an Adaptive Dynamic Search (ADS) algorithm that optimizes the order of Mel-Frequency Cepstral Coefficients (MFCCs) extracted from electroencephalogram (EEG) signals. This method strikes a balance between feature accuracy and redundancy, outperforming conventional fixed-order approaches. Subsequently, we analyze synchronization using the FastDTW-based Music–EEG Feature Fusion (FMEFF) framework, which leverages FastDTW to align music and EEG feature dimensions, effectively identifying the optimal alignment path to characterize synchronization. The fused features are then employed to classify music enjoyment levels, with explainability enhanced by examining classification performance across specific brain regions. Experimental results demonstrate enjoyment-level recognition accuracies of 74.50% and 81.54% in healthy and depressed participants, respectively, with the frontal lobe yielding the highest accuracy. Further correlation analysis confirms a strong relationship between the extracted features and subjective enjoyment scores. These findings present new insights for personalized music therapy recommendations and highlight the promise of integrating music and EEG signals in affective computing. This work also contributes to EEG channel optimization strategies and lays a foundation for future mental health interventions using wearable EEG systems.
- 28 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. 17 (1) , 511-528
FMEFF Mechanism: A FastDTW-Based Music-EEG Feature Fusion Approach for Identifying Enjoyment Levels in Music Therapy
Research indicates that individuals who experience greater music enjoyment exhibit stronger brain–music synchronization, which is positively correlated with improved therapeutic outcomes in depression treatment. However, most existing studies on music therapy predominantly focus on emotional states, often overlooking enjoyment levels. To address this gap, we propose a novel fusion mechanism based on Fast Dynamic Time Warping (FastDTW) to objectively assess music enjoyment. First, we introduce an Adaptive Dynamic Search (ADS) algorithm that optimizes the order of Mel-Frequency Cepstral Coefficients (MFCCs) extracted from electroencephalogram (EEG) signals. This method strikes a balance between feature accuracy and redundancy, outperforming conventional fixed-order approaches. Subsequently, we analyze synchronization using the FastDTW-based Music–EEG Feature Fusion (FMEFF) framework, which leverages FastDTW to align music and EEG feature dimensions, effectively identifying the optimal alignment path to characterize synchronization. The fused features are then employed to classify music enjoyment levels, with explainability enhanced by examining classification performance across specific brain regions. Experimental results demonstrate enjoyment-level recognition accuracies of 74.50% and 81.54% in healthy and depressed participants, respectively, with the frontal lobe yielding the highest accuracy. Further correlation analysis confirms a strong relationship between the extracted features and subjective enjoyment scores. These findings present new insights for personalized music therapy recommendations and highlight the promise of integrating music and EEG signals in affective computing. This work also contributes to EEG channel optimization strategies and lays a foundation for future mental health interventions using wearable EEG systems.
- 6 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-20
An Onboard Executable Multitask Network Model for Bioradar-Based ECG Signal Reconstruction Using High-Fidelity DHD Signals
Bioradar-based Electrocardiogram (ECG) reconstruction shows great promise in replacing traditional contact-based ECG collection methods for non-contact, long-term healthcare applications, such as continuous cardiorespiratory disease monitoring and early warning systems. In this work, we first designed a practical bioradar system for high Signal-to-Noise Ratio (SNR) I/Q baseband signal acquisition, achieving excellent performance with background noise not exceeding 3.18 μVpp and SNRs ranging from 26 to 119 dB. Next, we developed a high-linearity arctangent demodulation method to extract high-fidelity Doppler Heartbeat Diagram (DHD) signals. Finally, we introduced a lightweight UNet-based multitask network model for ECG signal reconstruction, which demonstrated performance achieving Root Mean Squared Error (RMSE) values of 0.160 and 0.330, Root Mean Absolute Error (RMAE) values of 0.261 and 0.400, and Pearson Correlation Coefficient (PCC) values of 95.17% and 85.26% for two different datasets, respectively. This model is characterized by low computational complexity, with 7.04M parameters, Floating-Point Operations (FLOPs) of 889.16 M, real-time processing speed of 1.05 s per execution, and low power consumption of 379.5mW. Moreover, it requires just 29.13MB of Random Access Memory (RAM) and 10.49MB of Read-Only Memory (ROM), making it highly suitable for deployment in embedded systems. Experimental results from both a public dataset and our own dataset show that the proposed lightweight ECG reconstruction model, when combined with the designed high-fidelity DHD signal acquisition bioradar system, holds significant potential for non-contact healthcare and medical applications.
- 6 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-20
An Onboard Executable Multitask Network Model for Bioradar-Based ECG Signal Reconstruction Using High-Fidelity DHD Signals
Bioradar-based Electrocardiogram (ECG) reconstruction shows great promise in replacing traditional contact-based ECG collection methods for non-contact, long-term healthcare applications, such as continuous cardiorespiratory disease monitoring and early warning systems. In this work, we first designed a practical bioradar system for high Signal-to-Noise Ratio (SNR) I/Q baseband signal acquisition, achieving excellent performance with background noise not exceeding 3.18 μVpp and SNRs ranging from 26 to 119 dB. Next, we developed a high-linearity arctangent demodulation method to extract high-fidelity Doppler Heartbeat Diagram (DHD) signals. Finally, we introduced a lightweight UNet-based multitask network model for ECG signal reconstruction, which demonstrated performance achieving Root Mean Squared Error (RMSE) values of 0.160 and 0.330, Root Mean Absolute Error (RMAE) values of 0.261 and 0.400, and Pearson Correlation Coefficient (PCC) values of 95.17% and 85.26% for two different datasets, respectively. This model is characterized by low computational complexity, with 7.04M parameters, Floating-Point Operations (FLOPs) of 889.16 M, real-time processing speed of 1.05 s per execution, and low power consumption of 379.5mW. Moreover, it requires just 29.13MB of Random Access Memory (RAM) and 10.49MB of Read-Only Memory (ROM), making it highly suitable for deployment in embedded systems. Experimental results from both a public dataset and our own dataset show that the proposed lightweight ECG reconstruction model, when combined with the designed high-fidelity DHD signal acquisition bioradar system, holds significant potential for non-contact healthcare and medical applications.
- 1 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 74 (11) , 3696-3709
An On-Board Executable Pareto-Based Iterated Local Search Algorithm for Embedded Multi-Core Processor Task Scheduling
The advancement of wearable electronic technology has facilitated the integration of smart wearable devices into artificial intelligence (AI)-driven medical assisted diagnosis. Embedded multi-core processors (MPs) have gradually emerged as pivotal hardware components for smart wearable medical diagnostic devices due to their high performance and flexibility. However, embedded MPs face the challenge of balancing performance, power consumption, and load-balancing. In response, we introduce a Pareto-based iterated local search (PILS) algorithm for task scheduling, which systematically optimizes multiple objectives, alongside a task list model to reduce the dimension of the decision space and enhance scheduling performance. In addition, we present a two-stage discretization scheme to ensure that the proposed algorithm offers meaningful guidance throughout the scheduling process. Simulation and on-board testing results show that the proposed algorithm effectively optimizes energy consumption, task execution time, and load balancing in embedded MPs task scheduling, indicating the potential of the proposed algorithm in enhancing the performance of smart wearable medical diagnostic devices powered by embedded MPs.
- 1 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 74 (11) , 3696-3709
An On-Board Executable Pareto-Based Iterated Local Search Algorithm for Embedded Multi-Core Processor Task Scheduling
The advancement of wearable electronic technology has facilitated the integration of smart wearable devices into artificial intelligence (AI)-driven medical assisted diagnosis. Embedded multi-core processors (MPs) have gradually emerged as pivotal hardware components for smart wearable medical diagnostic devices due to their high performance and flexibility. However, embedded MPs face the challenge of balancing performance, power consumption, and load-balancing. In response, we introduce a Pareto-based iterated local search (PILS) algorithm for task scheduling, which systematically optimizes multiple objectives, alongside a task list model to reduce the dimension of the decision space and enhance scheduling performance. In addition, we present a two-stage discretization scheme to ensure that the proposed algorithm offers meaningful guidance throughout the scheduling process. Simulation and on-board testing results show that the proposed algorithm effectively optimizes energy consumption, task execution time, and load balancing in embedded MPs task scheduling, indicating the potential of the proposed algorithm in enhancing the performance of smart wearable medical diagnostic devices powered by embedded MPs.
- 1 June 2025
- Published by IOP Publishing in Journal of Neural Engineering
- Vol. 22 (3) , 036040
Identifying and predicting EEG microstates with sequence-to-sequence deep learning models for online applications
Objective. Electroencephalographic (EEG) microstates, as a non-invasive and high-temporal-resolution tool for analyzing time-space features of brain activity, have been validated and applied in various research domains. However, current methods for EEG microstate analysis rely on clustering algorithms, which require large-scale offline computations to obtain microstate labels and cluster centers. This offline approach is no longer sufficient for applications in cross-subject, cross-dataset, and multi-task scenarios. Approach. To address these limitations, we propose, for the first time, a novel sequence-to-sequence-based framework for microstate identification and prediction, enabling end-to-end online recognition and prediction from EEG signals to microstate labels. Specifically, we introduce a method for constructing training datasets for online identification and prediction, which includes microstate label calibration, EEG electrode mapping, and sequence data partitioning. We validate this approach using four different neural network models with varying computational mechanisms on two public datasets. Main results. Our results demonstrate that EEG microstates can be identified and predicted by trainable models. In cross-subject microstate recognition tasks, the recognition accuracy for four typical microstates reached up to 74.26%, outperforming k-nearest neighbor (KNN) by 21.91%. For seven typical microstates, the recognition accuracy peaked at 66.76%, surpassing KNN by 26.6%. In prediction tasks, the accuracy for four and seven typical microstates reached 70.49% and 62.71%, respectively. Significance. Our work advances EEG microstate analysis from an offline clustering-based paradigm to an online model-data hybrid computation paradigm, providing new insights and references for cross-subject and cross-dataset applications of EEG microstates.
- 1 June 2025
- Published by IOP Publishing in Journal of Neural Engineering
- Vol. 22 (3) , 036040
Identifying and predicting EEG microstates with sequence-to-sequence deep learning models for online applications
Objective. Electroencephalographic (EEG) microstates, as a non-invasive and high-temporal-resolution tool for analyzing time-space features of brain activity, have been validated and applied in various research domains. However, current methods for EEG microstate analysis rely on clustering algorithms, which require large-scale offline computations to obtain microstate labels and cluster centers. This offline approach is no longer sufficient for applications in cross-subject, cross-dataset, and multi-task scenarios. Approach. To address these limitations, we propose, for the first time, a novel sequence-to-sequence-based framework for microstate identification and prediction, enabling end-to-end online recognition and prediction from EEG signals to microstate labels. Specifically, we introduce a method for constructing training datasets for online identification and prediction, which includes microstate label calibration, EEG electrode mapping, and sequence data partitioning. We validate this approach using four different neural network models with varying computational mechanisms on two public datasets. Main results. Our results demonstrate that EEG microstates can be identified and predicted by trainable models. In cross-subject microstate recognition tasks, the recognition accuracy for four typical microstates reached up to 74.26%, outperforming k-nearest neighbor (KNN) by 21.91%. For seven typical microstates, the recognition accuracy peaked at 66.76%, surpassing KNN by 26.6%. In prediction tasks, the accuracy for four and seven typical microstates reached 70.49% and 62.71%, respectively. Significance. Our work advances EEG microstate analysis from an offline clustering-based paradigm to an online model-data hybrid computation paradigm, providing new insights and references for cross-subject and cross-dataset applications of EEG microstates.
- Open Access
- 1 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Systems and Rehabilitation Engineering
- Vol. 33, 3647-3661
E-DANN: An Enhanced Domain Adaptation Network for Audio-EEG Feature Decoupling in Explainable Depression Recognition
Given the significant global health burden caused by depression, numerous studies have utilized artificial intelligence techniques to objectively and automatically detect depression. However, existing research primarily focuses on improving the accuracy of depression recognition while overlooking the explainability of detection models and the evaluation of feature importance. In this paper, we propose a novel framework named Enhanced Domain Adversarial Neural Network (E-DANN) for depression detection. First, we extract joint features that combine audio-specific physical properties and EEG responses to construct a multimodal feature space, which facilitates the understanding of how EEG signals dynamically respond to the physical properties of audio. Next, we employ the feature decoupling framework of E-DANN, which separates the extracted feature space into shared features and private features through adversarial training. The decoupled private features are then utilized for the binary classification of depression. Our experimental results validate the effectiveness of the proposed framework, which achieves accurate classification of normal controls and individuals with depression (accuracy: 92.28 ± 7.45%, specificity: 94.67 ± 7.30%, sensitivity: 89.11 ± 1.01%, and F1 score: 90.89 ± 8.60%). Furthermore, we employ an Explainable Artificial Intelligence (XAI) approach to hierarchically visualize feature importance and elucidate complex feature interaction patterns. In summary, this study provides a theoretical foundation for developing explainable diagnostic tools for depression and contributes to improving the clinical trustworthiness of AI-assisted diagnostic systems.
- 1 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Systems and Rehabilitation Engineering
- Vol. 33, 3647-3661
Open AccessE-DANN: An Enhanced Domain Adaptation Network for Audio-EEG Feature Decoupling in Explainable Depression Recognition
Given the significant global health burden caused by depression, numerous studies have utilized artificial intelligence techniques to objectively and automatically detect depression. However, existing research primarily focuses on improving the accuracy of depression recognition while overlooking the explainability of detection models and the evaluation of feature importance. In this paper, we propose a novel framework named Enhanced Domain Adversarial Neural Network (E-DANN) for depression detection. First, we extract joint features that combine audio-specific physical properties and EEG responses to construct a multimodal feature space, which facilitates the understanding of how EEG signals dynamically respond to the physical properties of audio. Next, we employ the feature decoupling framework of E-DANN, which separates the extracted feature space into shared features and private features through adversarial training. The decoupled private features are then utilized for the binary classification of depression. Our experimental results validate the effectiveness of the proposed framework, which achieves accurate classification of normal controls and individuals with depression (accuracy: 92.28 ± 7.45%, specificity: 94.67 ± 7.30%, sensitivity: 89.11 ± 1.01%, and F1 score: 90.89 ± 8.60%). Furthermore, we employ an Explainable Artificial Intelligence (XAI) approach to hierarchically visualize feature importance and elucidate complex feature interaction patterns. In summary, this study provides a theoretical foundation for developing explainable diagnostic tools for depression and contributes to improving the clinical trustworthiness of AI-assisted diagnostic systems.