A5031593162- Non-Invasive Vital Sign Monitoring
- IoT and Edge/Fog Computing
- Optical Imaging and Spectroscopy Techniques
- Context-Aware Activity Recognition Systems
- Gait Recognition and Analysis
- Scientific Computing and Data Management
- Video Surveillance and Tracking Methods
- Human Mobility and Location-Based Analysis
- Hemodynamic Monitoring and Therapy
- Anomaly Detection Techniques and Applications
- Data Quality and Management
- Heart Rate Variability and Autonomic Control
- Software-Defined Networks and 5G
- Advanced Data Storage Technologies
- Medical Imaging and Analysis
- Healthcare Technology and Patient Monitoring
- Remote Sensing and LiDAR Applications
- 3D Modeling in Geospatial Applications
- Human Pose and Action Recognition
- Advanced SAR Imaging Techniques
- Recommender Systems and Techniques
- Privacy-Preserving Technologies in Data
- Indoor and Outdoor Localization Technologies
- Urban and Freight Transport Logistics
- Data Management and Algorithms
Chinese Academy of Sciences
2021-2024
Institute of Computing Technology
2021-2024
University of Chinese Academy of Sciences
2017-2023
National University of Singapore
2019-2020
Zhejiang University
2020
Institute of Microelectronics
2017
With 5G on the verge of being adopted as next mobile network, there is a need to analyze its impact landscape computing and data management. In this paper, we both traditional emerging technologies project our view future research challenges opportunities. predicted increase 10-100× in bandwidth 5-10x decrease latency, expected be main enabler for smart cities, IoT efficient healthcare, where machine learning conducted at edge. context, investigate how can help development federated...
Training deep learning models for photoplethysmography(PPG)-based cuff-less blood pressure estimation often requires a substantial amount of labeled data collected through sophisticated medical instruments, posing significant challenges in practical applications. To address this issue, we propose Physiological Knowledge-Aware Contrastive Learning (PhysCL), novel approach designed to reduce the dependence on PPG while improving accuracy. Specifically, PhysCL tackles semantic consistency...
With the ever-increasing adoption of machine learning for data analytics, maintaining a pipeline is becoming more complex as both datasets and trained models evolve with time. In collaborative environment, changes updates due to evolution often cause cumbersome coordination maintenance work, raising costs making it hard use. Existing solutions, unfortunately, do not address version problem, especially in environment where non-linear control semantics are necessary isolate operations made by...
Deep learning-based methods demonstrate promising results in continuous non-invasive blood pressure measurement, whereas those models trained on large public datasets suffer from severe performance degradation predicting real-world user data collected home settings. Transfer learning has been recently introduced to personalize the pre-trained model with unseen users' solve problem. However, existing based network fine-tuning for personalization require a amount of labeled data, lacking...
The accuracy of noninvasive oxygen saturation (SpO2), which is defined by the measurements based on photoplethysmographic (PPG) signals, intensively affected motion artifacts (MAs) and low perfusion. This study introduces a novel approach called ESPRIT-MLT to measure SpO2 when such interferences are present. In contrast previous studies, work focuses harmonic model PPG signal probability results from analysis. optimized parametric ESPRIT method applied improve power estimation, maximum...
Social activities play an important role in people's daily life since they interact. For recommendations based on social activities, it is vital to have not only the activity information but also individuals' relations. Thanks geo-social networks and widespread use of location-aware mobile devices, massive data now readily available for exploitation by recommendation system. In this paper, a novel group method, called attentive recommendation, proposed recommend target user with both...
For low-semantic sensor signals from human activity recognition (HAR), contrastive learning (CL) is essential to implement novel applications or generic models without manual annotation, which a high-performance self-supervised (SSL) method. However, CL relies heavily on data augmentation for pairwise comparisons. Especially low semantic in the HAR area, conducting good performance strategies pretext tasks still rely attempts lacking generalizability and flexibility. To reduce burden, we...
Radar-based human activity recognition (HAR) is a time-series classification challenge, which widely applied in intelligent healthcare, ubiquitous computing, autonomous driving, etc. However, the received radar signal much noisier and less intuitive than images, makes it harder to capture posture information achieve high accuracy. In this paper, we use array collect data from different subjects propose novel simultaneous spatial-temporal encoding network (SST-HAR) extract features raw...
With 5G on the verge of being adopted as next mobile network, there is a need to analyze its impact landscape computing and data management. In this paper, we both traditional emerging technologies project our view future research challenges opportunities. predicted increase 10-100x in bandwidth 5-10x decrease latency, expected be main enabler for smart cities, IoT efficient healthcare, where machine learning conducted at edge. context, investigate how can help development federated...
Urban lifestyle depends on public services and retails, of which site locations matter to convenience for residents. We introduce a novel approach the systematic multi-site selection retails in an urban context. It takes as input set data about area generates optimal configuration two-dimensional sites retails. achieve this goal using data-driven optimisation entangling deep learning. The proposed can cost-efficiently generate location plan considering representative criteria, including...
With the ever-increasing adoption of machine learning for data analytics, maintaining a pipeline is becoming more complex as both datasets and trained models evolve with time. In collaborative environment, changes updates due to evolution often cause cumbersome coordination maintenance work, raising costs making it hard use. Existing solutions, unfortunately, do not address version problem, especially in environment where non-linear control semantics are necessary isolate operations made by...
Professional respiration monitors, essential for diagnosing physiological and psychological diseases, are expensive uncomfortable to wear. The inertial measurement unit (IMU) sensors on smart glasses can detect weak head movement caused by breathing, making them a more affordable wearable alternative. However, existing methods mainly evaluated in controlled environments with little motion artifacts cannot reconstruct the respiratory signal. In this paper, an encoder-decoder block...
Identifying human position in through-wall scenarios is quite helpful for emergency rescue, while the multiple reflections between wall and radar signal substantially degrade detection accuracy scenario, making it challenging to determine human's position. In this paper, we propose a novel approach trajectory tracking with radar. The group convolution mechanism used extract features of different views. And soft-squeeze excitation (SSE) module proposed fuse multi-view features, which...