Zihao Zhao
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
Zihao Zhao has a total of 77 co-authors. Here are the top co-authors by number of shared publications:
- Wenbo Ding
Tsinghua University
6 shared publications - Yang Liu
Hong Kong Polytechnic University, Hong Kong, China and Shanghai Artificial Intelligence Laboratory, Shanghai, China
3 shared publications Chengyue Lu
2 shared publications- Jiyu Wang
Huazhong University of Science and Technology
2 shared publications - Tian Lan
George Washington University, Washington, DC, USA
2 shared publications - Xiao-Ping Zhang
Tsinghua University
2 shared publications
Latest Publications
- Open Access
- 5 September 2025
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 11 (36) , eady0534
Wireless, wearable elastography via mechano-acoustic wave sensing for ambulatory monitoring of tissue stiffness
Assessing the mechanical properties of soft tissues holds broad clinical relevance. Advances in flexible electronics offer possibilities for wearable monitoring of tissue stiffness. However, existing technologies often rely on tethered setups or require frequent calibration, restricting their use in ambulatory environments. This study introduces a mechano-acoustic wave sensing technology for automated, wireless elastography. The patch-form sensor maintains conformal contact with the skin, regardless of body motion or deformation. It provides continuous, depth-sensitive estimation of subcutaneous tissue stiffness through real-time surface wave dispersion analysis. Theoretical and experimental investigations on phantom materials and tissues spanning a wide range of Young’s modulus (in kilopascals to megapascals) demonstrate the capability of the device to rapidly and robustly evaluate the stiffness at depths up to several centimeters. The device shows compatibility with various tissue models, with results consistent with in-parallel ultrasound elastography measurements. Deployment of the device during exercises confirms its viability for ambulatory monitoring, enabling continuous assessment of variation in tissue stiffness.
- 5 September 2025
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 11 (36) , eady0534
Open AccessWireless, wearable elastography via mechano-acoustic wave sensing for ambulatory monitoring of tissue stiffness
Assessing the mechanical properties of soft tissues holds broad clinical relevance. Advances in flexible electronics offer possibilities for wearable monitoring of tissue stiffness. However, existing technologies often rely on tethered setups or require frequent calibration, restricting their use in ambulatory environments. This study introduces a mechano-acoustic wave sensing technology for automated, wireless elastography. The patch-form sensor maintains conformal contact with the skin, regardless of body motion or deformation. It provides continuous, depth-sensitive estimation of subcutaneous tissue stiffness through real-time surface wave dispersion analysis. Theoretical and experimental investigations on phantom materials and tissues spanning a wide range of Young’s modulus (in kilopascals to megapascals) demonstrate the capability of the device to rapidly and robustly evaluate the stiffness at depths up to several centimeters. The device shows compatibility with various tissue models, with results consistent with in-parallel ultrasound elastography measurements. Deployment of the device during exercises confirms its viability for ambulatory monitoring, enabling continuous assessment of variation in tissue stiffness.
- Open Access
- 21 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 3867-3871
Beyond Aggregation: Efficient Federated Model Consolidation with Heterogeneity-Adaptive Weights Diffusion
As the Internet of Things (IoT) evolves, the need for enhanced data-sharing to improve edge device performance has led to the adoption of Federated Learning (FL) for data privacy and optimized data utilization. However, communication costs in FL remain a significant challenge. Traditional methods focus on client enhancements but overlook server-side aggregation, potentially increasing client computation loads. In response, we introduce a novel method, FedDiff, which utilizes diffusion models for generating model weights on FL servers, replacing traditional aggregation methods. Our approach, tailored for heterogeneous environments, significantly improves communication efficiency, achieving faster convergence and robust performance against weight noise in rigorous tests.
- 21 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 3867-3871
Open AccessBeyond Aggregation: Efficient Federated Model Consolidation with Heterogeneity-Adaptive Weights Diffusion
As the Internet of Things (IoT) evolves, the need for enhanced data-sharing to improve edge device performance has led to the adoption of Federated Learning (FL) for data privacy and optimized data utilization. However, communication costs in FL remain a significant challenge. Traditional methods focus on client enhancements but overlook server-side aggregation, potentially increasing client computation loads. In response, we introduce a novel method, FedDiff, which utilizes diffusion models for generating model weights on FL servers, replacing traditional aggregation methods. Our approach, tailored for heterogeneous environments, significantly improves communication efficiency, achieving faster convergence and robust performance against weight noise in rigorous tests.
- 15 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 23 (6) , 7363-7376
AQUILA: Communication Efficient Federated Learning With Adaptive Quantization in Device Selection Strategy
The widespread adoption of Federated Learning (FL), a privacy-preserving distributed learning methodology, has been impeded by the challenge of high communication overheads, typically arising from the transmission of large-scale models. Existing adaptive quantization methods, designed to mitigate these overheads, operate under the impractical assumption of uniform device participation. Additionally, these methods are limited in their adaptability due to the necessity of manual quantization level selection and often overlook biases inherent in local devices’ data, thereby affecting the robustness of the global model. In response, this paper introduces AQUILA ( a daptive qu antization in dev i ce se l ection str a tegy), a novel adaptive framework devised to effectively handle these issues, enhancing the efficiency and robustness of FL. AQUILA integrates a sophisticated device selection method that prioritizes the quality and usefulness of device updates. Utilizing the exact global model stored by devices enables a more precise device selection criterion, reduces model deviation, and limits the need for hyperparameter adjustments. Furthermore, AQUILA presents an innovative quantization criterion, optimized to improve communication efficiency while assuring model convergence. Our experiments demonstrate that AQUILA significantly decreases communication costs compared to existing methods, while maintaining comparable model performance across diverse non-homogeneous FL settings, such as Non-IID data and heterogeneous model architectures.
- 15 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 23 (6) , 7363-7376
AQUILA: Communication Efficient Federated Learning With Adaptive Quantization in Device Selection Strategy
The widespread adoption of Federated Learning (FL), a privacy-preserving distributed learning methodology, has been impeded by the challenge of high communication overheads, typically arising from the transmission of large-scale models. Existing adaptive quantization methods, designed to mitigate these overheads, operate under the impractical assumption of uniform device participation. Additionally, these methods are limited in their adaptability due to the necessity of manual quantization level selection and often overlook biases inherent in local devices’ data, thereby affecting the robustness of the global model. In response, this paper introduces AQUILA ( a daptive qu antization in dev i ce se l ection str a tegy), a novel adaptive framework devised to effectively handle these issues, enhancing the efficiency and robustness of FL. AQUILA integrates a sophisticated device selection method that prioritizes the quality and usefulness of device updates. Utilizing the exact global model stored by devices enables a more precise device selection criterion, reduces model deviation, and limits the need for hyperparameter adjustments. Furthermore, AQUILA presents an innovative quantization criterion, optimized to improve communication efficiency while assuring model convergence. Our experiments demonstrate that AQUILA significantly decreases communication costs compared to existing methods, while maintaining comparable model performance across diverse non-homogeneous FL settings, such as Non-IID data and heterogeneous model architectures.
- Open Access
- 8 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 724-729
Inclusive Data Representation in Federated Learning: A Novel Approach Integrating Textual and Visual Prompt
- 8 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 724-729
Open AccessInclusive Data Representation in Federated Learning: A Novel Approach Integrating Textual and Visual Prompt
- 18 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 23 (5) , 4819-4831
SAFARI: Sparsity-Enabled Federated Learning With Limited and Unreliable Communications
Federated learning (FL) enables edge devices to collaboratively learn a model in a distributed fashion. Many existing researches have focused on improving communication efficiency of high-dimensional models and addressing bias caused by local updates. However, most FL algorithms are either based on reliable communications or assuming fixed and known unreliability characteristics. In practice, networks could suffer from dynamic channel conditions and non-deterministic disruptions, with time-varying and unknown characteristics. To this end, in this paper we propose a sparsity-enabled FL framework with both improved communication efficiency and bias reduction, termed as SAFARI. It makes use of similarity among client models to rectify and compensate for bias that results from unreliable communications. More precisely, sparse learning is implemented on local clients to mitigate communication overhead, while to cope with unreliable communications, a similarity-based compensation method is proposed to provide surrogates for missing model updates. With respect to sparse models, we analyze SAFARI under bounded dissimilarity. It is demonstrated that SAFARI under unreliable communications is guaranteed to converge at the same rate as the standard FedAvg with perfect communications. Implementations and evaluations on the CIFAR-10 dataset validate the effectiveness of SAFARI by showing that it can achieve the same convergence speed and accuracy as FedAvg with perfect communications, with up to 60% of the model weights being pruned and a high percentage of client updates missing in each round of model updates.
- 18 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 23 (5) , 4819-4831
SAFARI: Sparsity-Enabled Federated Learning With Limited and Unreliable Communications
Federated learning (FL) enables edge devices to collaboratively learn a model in a distributed fashion. Many existing researches have focused on improving communication efficiency of high-dimensional models and addressing bias caused by local updates. However, most FL algorithms are either based on reliable communications or assuming fixed and known unreliability characteristics. In practice, networks could suffer from dynamic channel conditions and non-deterministic disruptions, with time-varying and unknown characteristics. To this end, in this paper we propose a sparsity-enabled FL framework with both improved communication efficiency and bias reduction, termed as SAFARI. It makes use of similarity among client models to rectify and compensate for bias that results from unreliable communications. More precisely, sparse learning is implemented on local clients to mitigate communication overhead, while to cope with unreliable communications, a similarity-based compensation method is proposed to provide surrogates for missing model updates. With respect to sparse models, we analyze SAFARI under bounded dissimilarity. It is demonstrated that SAFARI under unreliable communications is guaranteed to converge at the same rate as the standard FedAvg with perfect communications. Implementations and evaluations on the CIFAR-10 dataset validate the effectiveness of SAFARI by showing that it can achieve the same convergence speed and accuracy as FedAvg with perfect communications, with up to 60% of the model weights being pruned and a high percentage of client updates missing in each round of model updates.
- 20 April 2023
- Published by Elsevier in Nano Energy
- Vol. 112
A triboelectric gait sensor system for human activity recognition and user identification
- 20 April 2023
- Published by Elsevier in Nano Energy
- Vol. 112
A triboelectric gait sensor system for human activity recognition and user identification
- Top Cited Papers
- 6 January 2023
- Published by Elsevier in Journal of the Franklin Institute
- Vol. 360 (12) , 8669-8703
Towards efficient communications in federated learning: A contemporary survey
- 6 January 2023
- Published by Elsevier in Journal of the Franklin Institute
- Vol. 360 (12) , 8669-8703
Top Cited PapersTowards efficient communications in federated learning: A contemporary survey
- Open Access
- 1 January 2021
- Published by Elsevier in SSRN Electronic Journal
A Flexible Triboelectric Tactile Sensor for Simultaneous Material and Texture Recognition
Electronic skin with tactile perception enables intelligent robots and prostheses to perform dexterous manipulation and natural interaction with the human and surroundings. However, using single tactile sensing mechanism to simultaneously percept geometry features and materials properties remains a challenge due to the bottleneck of signal decoupling. Herein, we report the MTSensing system – a wireless and fully-integrated tactile sensing system that can simultaneously recognize materials and textures based on a single flexible triboelectric sensor. The proposed triboelectric sensor converts touch into electrical signals and meanwhile, the signal processing pipeline decouples the signals into macro/micro features and feeds them into the corresponding deep learning models, which simultaneously predict the materials and textures of the contacted objects with the accuracies of 99.07% and 99.32%, respectively. The systematic integration of MTSensing hopes to pave the way for deploying low-cost and scalable electronic skin with multi-functional perceptions.
- 1 January 2021
- Published by Elsevier in SSRN Electronic Journal
Open AccessA Flexible Triboelectric Tactile Sensor for Simultaneous Material and Texture Recognition
Electronic skin with tactile perception enables intelligent robots and prostheses to perform dexterous manipulation and natural interaction with the human and surroundings. However, using single tactile sensing mechanism to simultaneously percept geometry features and materials properties remains a challenge due to the bottleneck of signal decoupling. Herein, we report the MTSensing system – a wireless and fully-integrated tactile sensing system that can simultaneously recognize materials and textures based on a single flexible triboelectric sensor. The proposed triboelectric sensor converts touch into electrical signals and meanwhile, the signal processing pipeline decouples the signals into macro/micro features and feeds them into the corresponding deep learning models, which simultaneously predict the materials and textures of the contacted objects with the accuracies of 99.07% and 99.32%, respectively. The systematic integration of MTSensing hopes to pave the way for deploying low-cost and scalable electronic skin with multi-functional perceptions.
- 18 December 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 18 (1) , 467-476
MAGLeak: A Learning-Based Side-Channel Attack for Password Recognition With Multiple Sensors in IIoT Environment
As an emerging technology, industrial Internet of Things (IIoT) connects massive sensors and actuators to empower industrial sectors being smart, autonomous, efficient, and safety. However, due the large number of build-in sensors of IIoT smart devices, the IIoT systems are vulnerable to side-channel attack. In this article, a novel side-channel-based passwords cracking system, namely MAGLeak, is proposed to recognize the victim's passwords by leveraging accelerometer, gyroscope, and magnetometer of IIoT touch-screen smart device. Specifically, an event-driven data collection method is proposed to ensure that the user's keystroke behavior can be reflected accurately by the obtained measurements of three sensors. Moreover, random forest algorithm is leveraged for the recognition module, followed by a data preprocessing process. Extensive experimental results demonstrate that MAGLeak achieves a high recognition accuracy under small training dataset, e.g., achieving recognition accuracy 98% of each single key for 2000 training samples.
- 18 December 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 18 (1) , 467-476
MAGLeak: A Learning-Based Side-Channel Attack for Password Recognition With Multiple Sensors in IIoT Environment
As an emerging technology, industrial Internet of Things (IIoT) connects massive sensors and actuators to empower industrial sectors being smart, autonomous, efficient, and safety. However, due the large number of build-in sensors of IIoT smart devices, the IIoT systems are vulnerable to side-channel attack. In this article, a novel side-channel-based passwords cracking system, namely MAGLeak, is proposed to recognize the victim's passwords by leveraging accelerometer, gyroscope, and magnetometer of IIoT touch-screen smart device. Specifically, an event-driven data collection method is proposed to ensure that the user's keystroke behavior can be reflected accurately by the obtained measurements of three sensors. Moreover, random forest algorithm is leveraged for the recognition module, followed by a data preprocessing process. Extensive experimental results demonstrate that MAGLeak achieves a high recognition accuracy under small training dataset, e.g., achieving recognition accuracy 98% of each single key for 2000 training samples.
- RetractedOpen Access
- 1 December 2020
- Published by EDP Sciences in E3S Web of Conferences
- Vol. 213, 02030
Research on Key Technologies and Scenario Application of Emerging Business under the Energy Internet
Under the energy Internet situation, the market-oriented competition of emerging power grid businesses is increasingly intensified. The traditional single-type information construction system and model cannot adapt to the market. Therefore, this paper proposes a shared service architecture technology for China and Taiwan. This article introduces the goals and architectural ideas of the China-Taiwan shared service architecture and proposes a five-step method to sort out shared service centers. When the enterprise implements the China-Taiwan shared service center architecture, it proposes the implementation strategy of minimizing closed-loop principles and staged iterations, which are shown through research and application., China-Taiwan shared service architecture can support grid companies to quickly respond to the needs of emerging power business markets, and play an IT support role for grid companies in the transformation process.
- 1 December 2020
- Published by EDP Sciences in E3S Web of Conferences
- Vol. 213, 02030
RetractedOpen AccessResearch on Key Technologies and Scenario Application of Emerging Business under the Energy Internet
Under the energy Internet situation, the market-oriented competition of emerging power grid businesses is increasingly intensified. The traditional single-type information construction system and model cannot adapt to the market. Therefore, this paper proposes a shared service architecture technology for China and Taiwan. This article introduces the goals and architectural ideas of the China-Taiwan shared service architecture and proposes a five-step method to sort out shared service centers. When the enterprise implements the China-Taiwan shared service center architecture, it proposes the implementation strategy of minimizing closed-loop principles and staged iterations, which are shown through research and application., China-Taiwan shared service architecture can support grid companies to quickly respond to the needs of emerging power business markets, and play an IT support role for grid companies in the transformation process.