A5100658136- Wireless Body Area Networks
- BIM and Construction Integration
- Educational Systems and Policies
- Advanced Neural Network Applications
- 3D IC and TSV technologies
- Educational Research and Pedagogy
- Energy Harvesting in Wireless Networks
- Advanced Sensor and Energy Harvesting Materials
- Infrastructure Maintenance and Monitoring
- Wireless Power Transfer Systems
- Context-Aware Activity Recognition Systems
- Education and Learning Interventions
- Fault Detection and Control Systems
- Occupational Health and Safety Research
- Advanced biosensing and bioanalysis techniques
- Muscle activation and electromyography studies
- Nanoplatforms for cancer theranostics
- Radio Frequency Integrated Circuit Design
- IoT and Edge/Fog Computing
- Laser-Ablation Synthesis of Nanoparticles
- ZnO doping and properties
- Construction Project Management and Performance
- Green IT and Sustainability
- Innovative Energy Harvesting Technologies
- Manufacturing Process and Optimization
Kwangwoon University
2010-2025
Ulsan National Institute of Science and Technology
2022-2024
University of North Carolina at Chapel Hill
2018-2021
Korea University
2017-2019
Gyeongsang National University
2011-2018
Jeonbuk National University
2016
Korea Advanced Institute of Science and Technology
2008-2014
Imec the Netherlands
2014
National Institutes of Health
2010-2013
Holst Centre (Netherlands)
2013
A self-configured body sensor network controller and a high efficiency wirelessly powered are presented for wearable, continuous health monitoring system. The chip harvests its power from the surrounding band using an Adaptive Threshold Rectifier (ATR) with 54.9% efficiency, it consumes 12 μW to implement electrocardiogram (ECG) analog front-end ADC. ATR is implemented standard CMOS process low cost. adhesive bandage type patch composed of chip, Planar-Fashionable Circuit Board (P-FCB)...
A wearable electrocardiogram (ECG) acquisition system implemented with planar-fashionable circuit board (P-FCB)-based shirt is presented. The proposed removes cumbersome wires from conventional Holter monitor for convenience. Dry electrodes screen-printed directly on fabric enables long-term monitoring without skin irritation. ECG exploits a chip group of around the body, and both interconnection are using P-FCB to enhance wearability lower production cost. characteristics electrode shown,...
A low power highly sensitive Thoracic Impedance Variance (TIV) and Electrocardiogram (ECG) monitoring SoC is designed implemented into a poultice-like plaster sensor for wearable cardiac monitoring. 0.1 Ω TIV detection possible with sensitivity of 3.17 V/Ω SNR > 40 dB. This achieved the help high quality (Q-factor 30) balanced sinusoidal current source noise reconfigurable readout electronics. cm-range 13.56 MHz fabric inductor coupling adopted to start/stop remotely. Moreover, 5%...
Various imaging modalities have been exploited and are successfully employed to investigate the anatomic or functional dissemination of tissues in body. However, no single modality allows overall structural, functional, molecular information as each has its own unique strengths weaknesses. The combination two popular that investigates different methods might offer prospect improved diagnostic therapeutic monitoring abilities. To date, only a few hybrid systems become clinically adopted, but...
Recently, several attempts have been made to continuously monitor chronic diseases in everyday life, but their large form factors or battery power limitations still remain be solved. Security and/or requirements of interference resilience are stringent for health monitoring, and therefore, using an open-air wireless ISM band [1] is not suitable a monitoring body sensor network (BSN). This paper presents self-configured wearable BSN system with high efficiency wirelessly powered sensors that...
This paper introduces intermittent learning --- the goal of which is to enable energy harvested computing platforms capable executing certain classes machine tasks effectively and efficiently. We identify unique challenges relating data application semantics tasks, address these challenges, we devise 1) an algorithm that determines a sequence actions achieve desired objective under tight constraints, 2) propose three heuristics help learner decide whether learn or discard training examples...
We introduce SubFlow-a dynamic adaptation and execution strategy for a deep neural network (DNN), which enables real-time DNN inference training. The goal of SubFlow is to complete the task within timing constraint that may dynamically change while ensuring comparable performance executing full by subset at run-time. To this end, we propose two online algorithms enable SubFlow: 1) construction sub-network constructs best subnetwork in terms size configuration, 2) time-bound executes given...
We introduce Neuro.ZERO---a co-processor architecture consisting of a main microcontroller (MCU) that executes scaled-down versions deep neural network1 (DNN) inference task, and an accelerator is powered by harvested energy follows the intermittent computing paradigm [76]. The goal to enhance performance DNN running on microcontroller. Neuro.ZERO opportunistically accelerates run-time via one its four acceleration modes: extended inference, expedited ensemble latent training. To enable...
This paper introduces the concept of Neural Weight Virtualization - which enables fast and scalable in-memory multitask deep learning on memory-constrained embedded systems. The goal neural weight virtualization is two-fold: (1) packing multiple DNNs into a fixed-sized main memory whose combined requirement larger than memory, (2) enabling execution DNNs. To this end, we propose two-phase approach: parameters for fine-grained parameter sharing at level weights that scales up to heterogeneous...
We propose SMMF (Square-Matricized Momentum Factorization), a memory-efficient optimizer that reduces the memory requirement of widely used adaptive learning rate optimizers, such as Adam, by up to 96%. enables flexible and efficient factorization an arbitrary rank (shape) first second momentum tensors during optimization, based on proposed square-matricization one-time single matrix factorization. From this, it becomes effectively applicable any tensors, i.e., bias, matrix, rank-d prevalent...
A new flexible electronics technology, named P-FCB (Planar Fashionable Circuit Board), is introduced. a circuit board technology implemented on the plain fabric patch for wearable applications. In this paper, manufacturing of PFCB, and its electrical characteristics such as sheet resistance, maximum current density, frequency are reported. The fabrication methods their passive devices resistor, capacitor, inductor in discussed. addition, how to integrate silicon chip directly system...
A near-field coupling transceiver integrated with a fault-tolerant network switch is implemented for inter-layer and intra-layer wearable body area network. The inductive employs resonance compensator (RC) digitally controlled on-chip capacitor bank variable hysteresis Schmitt trigger to compensate dynamic static variances of woven inductor, it enables 10 Mbps wireless transaction the reception energy 1.12 pJ/b at 2.5 V supply. introduces new protocol eliminate routing table reduces power...
A light-weighted body area network using 3-layer coin-sized fabric patches is proposed for sleep monitoring systems. It consists of two ICs, a controller (NC) and sensor node (SN), also Wearable Band (W-Band) to connect them. The Continuous Data Transmission (CDT) protocol low power real-time scalability by in-order data transmission W-Band among several sensors. Based on this protocol, Linked List based Manager (LLM) Adaptive Dual Mode Controller (ADMC) in NC, swing transmitter (D-TX) its...
Implantable and ambulatory measurement of physiological signals such as Bio-impedance using miniature biomedical devices needs careful tradeoff between limited power budget, accuracy complexity implementation. This paper addresses this through an extensive analysis different stimulation demodulation techniques for accurate measurement. Three cases are considered rigorous a generic impedance model, with multiple poles, which is stimulated square/sinusoidal current demodulated clock. For each...