Electrical+Engineering
-
KAIST Presents Game-Changing Technology for Intractable Brain Disease Treatment Using Micro OLEDs
<(From left)Professor Kyung Cheol Choi, Hyunjoo J. Lee, Somin Lee from the School of Electrical Engineering>
Optogenetics is a technique that controls neural activity by stimulating neurons expressing light-sensitive proteins with specific wavelengths of light. It has opened new possibilities for identifying causes of brain disorders and developing treatments for intractable neurological diseases. Because this technology requires precise stimulation inside the human brain with minimal damage to soft brain tissue, it must be integrated into a neural probe—a medical device implanted in the brain. KAIST researchers have now proposed a new paradigm for neural probes by integrating micro OLEDs into thin, flexible, implantable medical devices.
KAIST (President Kwang Hyung Lee) announced on the 6th of July that Professor Kyung Cheol Choi and researcher Hyunjoo J. Lee from the School of Electrical Engineering have jointly succeeded in developing an optogenetic neural probe integrated with flexible micro OLEDs.
Optical fibers have been used for decades in optogenetic research to deliver light to deep brain regions from external light sources. Recently, research has focused on flexible optical fibers and ultra-miniaturized neural probes that integrate light sources for single-neuron stimulation.
The research team focused on micro OLEDs due to their high spatial resolution and flexibility, which allow for precise light delivery to small areas of neurons. This enables detailed brain circuit analysis while minimizing side effects and avoiding restrictions on animal movement. Moreover, micro OLEDs offer precise control of light wavelengths and support multi-site stimulation, making them suitable for studying complex brain functions.
However, the device's electrical properties degrade easily in the presence of moisture or water, which limited their use as implantable bioelectronics. Furthermore, optimizing the high-resolution integration process on thin, flexible probes remained a challenge.
To address this, the team enhanced the operational reliability of OLEDs in moist, oxygen-rich environments and minimized tissue damage during implantation. They patterned an ultrathin, flexible encapsulation layer* composed of aluminum oxide and parylene-C (Al₂O₃/parylene-C) at widths of 260–600 micrometers (μm) to maintain biocompatibility.
*Encapsulation layer: A barrier that completely blocks oxygen and water molecules from the external environment, ensuring the longevity and reliability of the device.
When integrating the high-resolution micro OLEDs, the researchers also used parylene-C, the same biocompatible material as the encapsulation layer, to maintain flexibility and safety. To eliminate electrical interference between adjacent OLED pixels and spatially separate them, they introduced a pixel define layer (PDL), enabling the independent operation of eight micro OLEDs.
Furthermore, they precisely controlled the residual stress and thickness in the multilayer film structure of the device, ensuring its flexibility even in biological environments. This optimization allowed for probe insertion without bending or external shuttles or needles, minimizing mechanical stress during implantation.
Advanced Functional Materials-Conceptual diagram of a flexible neural probe for integrated optogenetics (Micro-OLED)>
As a result, the team developed a flexible neural probe with integrated micro OLEDs capable of emitting more than one milliwatt per square millimeter (mW/mm²) at 470 nanometers (nm), the optimal wavelength for activating channelrhodopsin-2. This is a significantly high light output for optogenetics and biomedical stimulation applications.
The ultrathin flexible encapsulation layer exhibited a low water vapor transmission rate of 2.66×10⁻⁵ g/m²/day, allowing the device to maintain functionality for over 10 years. The parylene-C-based barrier also demonstrated excellent performance in biological environments, successfully enabling the independent operation of the integrated OLEDs without electrical interference or bending issues.
Dr. Somin Lee, the lead author from Professor Choi’s lab, stated, “We focused on fine-tuning the integration process of highly flexible, high-resolution micro OLEDs onto thin flexible probes, enhancing their biocompatibility and application potential. This is the first reported development of such flexible OLEDs in a probe format and presents a new paradigm for using flexible OLEDs as implantable medical devices for monitoring and therapy.”
This study, with Dr. Somin Lee as the first author, was published online on March 26 in Advanced Functional Materials (IF 18.5), a leading international journal in the field of nanotechnology, and was selected as the cover article for the upcoming July issue.
※ Title: Advanced Micro-OLED Integration on Thin and Flexible Polymer Neural Probes for Targeted Optogenetic Stimulation ※ DOI: https://doi.org/10.1002/adfm.202420758
The research was supported by the Ministry of Science and ICT and the National Research Foundation of Korea through the Electronic Medicine Technology Development Program (Project title: Development of Core Source Technologies and In Vivo Validation for Brain Cognition and Emotion-Enhancing Light-Stimulating Electronic Medicine).
2025.07.07 View 70 -
KAIST researcher Se Jin Park develops 'SpeechSSM,' opening up possibilities for a 24-hour AI voice assistant.
<(From Left)Prof. Yong Man Ro and Ph.D. candidate Sejin Park>
Se Jin Park, a researcher from Professor Yong Man Ro’s team at KAIST, has announced 'SpeechSSM', a spoken language model capable of generating long-duration speech that sounds natural and remains consistent.
An efficient processing technique based on linear sequence modeling overcomes the limitations of existing spoken language models, enabling high-quality speech generation without time constraints.
It is expected to be widely used in podcasts, audiobooks, and voice assistants due to its ability to generate natural, long-duration speech like humans.
Recently, Spoken Language Models (SLMs) have been spotlighted as next-generation technology that surpasses the limitations of text-based language models by learning human speech without text to understand and generate linguistic and non-linguistic information. However, existing models showed significant limitations in generating long-duration content required for podcasts, audiobooks, and voice assistants. Now, KAIST researcher has succeeded in overcoming these limitations by developing 'SpeechSSM,' which enables consistent and natural speech generation without time constraints.
KAIST(President Kwang Hyung Lee) announced on the 3rd of July that Ph.D. candidate Sejin Park from Professor Yong Man Ro's research team in the School of Electrical Engineering has developed 'SpeechSSM,' a spoken. a spoken language model capable of generating long-duration speech.
This research is set to be presented as an oral paper at ICML (International Conference on Machine Learning) 2025, one of the top machine learning conferences, selected among approximately 1% of all submitted papers. This not only proves outstanding research ability but also serves as an opportunity to once again demonstrate KAIST's world-leading AI research capabilities.
A major advantage of Spoken Language Models (SLMs) is their ability to directly process speech without intermediate text conversion, leveraging the unique acoustic characteristics of human speakers, allowing for the rapid generation of high-quality speech even in large-scale models.
However, existing models faced difficulties in maintaining semantic and speaker consistency for long-duration speech due to increased 'speech token resolution' and memory consumption when capturing very detailed information by breaking down speech into fine fragments.
To solve this problem, Se Jin Park developed 'SpeechSSM,' a spoken language model using a Hybrid State-Space Model, designed to efficiently process and generate long speech sequences.
This model employs a 'hybrid structure' that alternately places 'attention layers' focusing on recent information and 'recurrent layers' that remember the overall narrative flow (long-term context). This allows the story to flow smoothly without losing coherence even when generating speech for a long time. Furthermore, memory usage and computational load do not increase sharply with input length, enabling stable and efficient learning and the generation of long-duration speech.
SpeechSSM effectively processes unbounded speech sequences by dividing speech data into short, fixed units (windows), processing each unit independently, and then combining them to create long speech.
Additionally, in the speech generation phase, it uses a 'Non-Autoregressive' audio synthesis model (SoundStorm), which rapidly generates multiple parts at once instead of slowly creating one character or one word at a time, enabling the fast generation of high-quality speech.
While existing models typically evaluated short speech models of about 10 seconds, Se Jin Park created new evaluation tasks for speech generation based on their self-built benchmark dataset, 'LibriSpeech-Long,' capable of generating up to 16 minutes of speech.
Compared to PPL (Perplexity), an existing speech model evaluation metric that only indicates grammatical correctness, she proposed new evaluation metrics such as 'SC-L (semantic coherence over time)' to assess content coherence over time, and 'N-MOS-T (naturalness mean opinion score over time)' to evaluate naturalness over time, enabling more effective and precise evaluation.
Through these new evaluations, it was confirmed that speech generated by the SpeechSSM spoken language model consistently featured specific individuals mentioned in the initial prompt, and new characters and events unfolded naturally and contextually consistently, despite long-duration generation. This contrasts sharply with existing models, which tended to easily lose their topic and exhibit repetition during long-duration generation.
PhD candidate Sejin Park explained, "Existing spoken language models had limitations in long-duration generation, so our goal was to develop a spoken language model capable of generating long-duration speech for actual human use." She added, "This research achievement is expected to greatly contribute to various types of voice content creation and voice AI fields like voice assistants, by maintaining consistent content in long contexts and responding more efficiently and quickly in real time than existing methods."
This research, with Se Jin Park as the first author, was conducted in collaboration with Google DeepMind and is scheduled to be presented as an oral presentation at ICML (International Conference on Machine Learning) 2025 on July 16th.
Paper Title: Long-Form Speech Generation with Spoken Language Models
DOI: 10.48550/arXiv.2412.18603
Ph.D. candidate Se Jin Park has demonstrated outstanding research capabilities as a member of Professor Yong Man Ro's MLLM (multimodal large language model) research team, through her work integrating vision, speech, and language. Her achievements include a spotlight paper presentation at 2024 CVPR (Computer Vision and Pattern Recognition) and an Outstanding Paper Award at 2024 ACL (Association for Computational Linguistics).
For more information, you can refer to the publication and accompanying demo: SpeechSSM Publications.
2025.07.04 View 407 -
Manipulating Brain Cells by Smartphone
Researchers have developed a soft neural implant that can be wirelessly controlled using a smartphone. It is the first wireless neural device capable of indefinitely delivering multiple drugs and multiple colour lights, which neuroscientists believe can speed up efforts to uncover brain diseases such as Parkinson’s, Alzheimer’s, addiction, depression, and pain.
A team under Professor Jae-Woong Jeong from the School of Electrical Engineering at KAIST and his collaborators have invented a device that can control neural circuits using a tiny brain implant controlled by a smartphone. The device, using Lego-like replaceable drug cartridges and powerful, low-energy Bluetooth, can target specific neurons of interest using drugs and light for prolonged periods. This study was published in Nature Biomedical Engineering.
“This novel device is the fruit of advanced electronics design and powerful micro and nanoscale engineering,” explained Professor Jeong. “We are interested in further developing this technology to make a brain implant for clinical applications.”
This technology significantly overshadows the conventional methods used by neuroscientists, which usually involve rigid metal tubes and optical fibers to deliver drugs and light. Apart from limiting the subject’s movement due to bulky equipment, their relatively rigid structure causes lesions in soft brain tissue over time, therefore making them not suitable for long-term implantation. Although some efforts have been made to partly mitigate adverse tissue response by incorporating soft probes and wireless platforms, the previous solutions were limited by their inability to deliver drugs for long periods of time as well as their bulky and complex control setups.
To achieve chronic wireless drug delivery, scientists had to solve the critical challenge of the exhaustion and evaporation of drugs. To combat this, the researchers invented a neural device with a replaceable drug cartridge, which could allow neuroscientists to study the same brain circuits for several months without worrying about running out of drugs.
These ‘plug-n-play’ drug cartridges were assembled into a brain implant for mice with a soft and ultrathin probe (with the thickness of a human hair), which consisted of microfluidic channels and tiny LEDs (smaller than a grain of salt), for unlimited drug doses and light delivery.
Controlled with an elegant and simple user interface on a smartphone, neuroscientists can easily trigger any specific combination or precise sequencing of light and drug delivery in any implanted target animal without the need to be physically inside the laboratory. Using these wireless neural devices, researchers can also easily setup fully automated animal studies where the behaviour of one animal could affect other animals by triggering light and/or drug delivery.
“The wireless neural device enables chronic chemical and optical neuromodulation that has never been achieved before,” said lead author Raza Qazi, a researcher with KAIST and the University of Colorado Boulder.
This work was supported by grants from the National Research Foundation of Korea, US National Institute of Health, National Institute on Drug Abuse, and Mallinckrodt Professorship.
(A neural implant with replaceable drug cartridges and Bluetooth low-energy can target specific neurons .)
(Micro LED controlling using smartphone application)
2019.08.07 View 34651 -
Anti-drone Technology for Anti-Terrorism Applications
(from top right clockwise: Professor Yongdae Kim,
PhD Candidates Yujin Kwon, Juhwan Noh, Hocheol Shin, and Dohyun Kim)
KAIST researchers have developed anti-drone technology that can hijack other drones by spoofing its location using fake GPS signals. This technology can safely guide a drone to a desired location without any sudden change in direction in emergency situations, and thus respond effectively to dangerous drones such as those intending to carry out acts of terrorism.
Advancements in the drone industry have led to the wider use of drones in our daily lives in areas of reconnaissance, searching and rescuing, disaster prevention and response, and delivery services. At the same time, there has also been a growing concern about privacy, safety, and security issues regarding drones, especially those arising from intrusion into private property and secure facilities. Therefore, the anti-drone industry is rapidly expanding to detect and respond to this possible drone invasion.
The current anti-drone systems used in airports and other key locations utilize electronic jamming signals, high-power lasers, or nets to neutralize drones. For example, drones trespassing on airports are often countered with simple jamming signals that can prevent the drones from moving and changing position, but this may result in a prolonged delay in flight departures and arrivals at the airports. Drones used for terrorist attacks – armed with explosives or weapons – must also be neutralized a safe distance from the public and vital infrastructure to minimize any damage.
Due to this need for a new anti-drone technology to counter these threats, a KAIST research team led by Professor Yongdae Kim from the School of Electrical Engineering has developed technology that securely thwarts drones by tricking them with fake GPS signals.
Fake GPS signals have been used in previous studies to cause confusion inside the drone regarding its location, making the drone drift from its position or path. However, such attack tactics cannot be applied in GPS safety mode. GPS safety mode is an emergency mode that ensures drone safety when the signal is cut or location accuracy is low due to a fake GPS signals. This mode differs between models and manufacturers.
Professor Kim’s team analyzed the GPS safety mode of different drone models made from major drone manufacturers such as DJI and Parrot, made classification systems, and designed a drone abduction technique that covers almost all the types of drone GPS safety modes, and is universally applicable to any drone that uses GPS regardless of model or manufacturer. The research team applied their new technique to four different drones and have proven that the drones can be safely hijacked and guided to the direction of intentional abduction within a small margin of error.
Professor Kim said, “Conventional consumer drones equipped with GPS safety mode seem to be safe from fake GPS signals, however, most of these drones are able to be detoured since they detect GPS errors in a rudimentary manner.” He continued, “This technology can contribute particularly to reducing damage to airports and the airline industry caused by illegal drone flights.”
The research team is planning to commercialize the developed technology by applying it to existing anti-drone solutions through technology transfer.” This research, featured in the ACM Transactions on Privacy and Security (TOPS) on April 9, was supported by the Defense Acquisition Program Administration (DAPA) and the Agency for Defense Development (ADD).
Image 1.
Experimental environment in which a fake GPS signal was produced from a PC and injected into the drone signal using directional antennae
Publication:
Juhwan Noh, Yujin Kwon, Yunmok Son, Hocheol Shin, Dohyun Kim, Jaeyeong Choi, and Yongdae Kim. 2019. Tractor Beam: Safe-hijacking of Consumer Drones with Adaptive GPS Spoofing. ACM Transactions on Privacy and Security. New York, NY, USA, Vol. 22, No. 2, Article 12, 26 pages. https://doi.org/10.1145/3309735
Profile: Prof. Yongdae Kim, MS, PhD
yongdaek@kaist.ac.kr
https://www.syssec.kr/
Professor
School of Electrical Engineering
Korea Advanced Institute of Science and Technology (KAIST)
http://kaist.ac.kr Daejeon 34141, Korea
Profile: Juhwan Noh, PhD Candidate
juhwan@kaist.ac.kr
PhD Candidate
System Security (SysSec) Lab
School of Electrical Engineering
Korea Advanced Institute of Science and Technology (KAIST)
http://kaist.ac.kr Daejeon 34141, Korea
(END)
2019.06.25 View 45878 -
True-meaning Wearable Displays: Self-powered, Washable and Wearable
(Video: The washing process of wearing display module) When we think about clothes, they are usually formed with textiles and have to be both wearable and washable for daily use; however, smart clothing has had a problem with its power sources and moisture permeability, which causes the devices to malfunction. This problem has now been overcome by a KAIST research team, who developed a textile-based wearable display module technology that is washable and does not require an external power source.
To ease out the problem of external power sources and enhance the practicability of wearable displays, Professor Kyung Cheol Choi from the School of Electrical Engineering and his team fabricated their wearing display modules on real textiles that integrated polymer solar cells (PSCs) with organic light emitting diodes (OLEDs).
PSCs have been one of the most promising candidates for a next-generation power source, especially for wearable and optoelectronic applications because they can provide stable power without an external power source, while OLEDs can be driven with milliwatts. However, the problem was that they are both very vulnerable to external moisture and oxygen. The encapsulation barrier is essential for their reliability. The conventional encapsulation barrier is sufficient for normal environments; however, it loses its characteristics in aqueous environments, such as water. It limits the commercialization of wearing displays that must operate even on rainy days or after washing.
To tackle this issue, the team employed a washable encapsulation barrier that can protect the device without losing its characteristics after washing through atomic layer deposition (ALD) and spin coating. With this encapsulation technology, the team confirmed that textile-based wearing display modules including PSCs, OLEDs, and the proposed encapsulation barrier exhibited little change in characteristics even after 20 washings with 10-minute cycles. Moreover, the encapsulated device operated stably with a low curvature radius of 3mm and boasted high reliability.
Finally, it exhibited no deterioration in properties over 30 days even after being subjected to both bending stress and washing. Since it uses a less stressful textile, compared to conventional wearable electronic devices that use traditional plastic substrates, this technology can accelerate the commercialization of wearing electronic devices. Importantly, this wearable electronic device in daily life can save energy through a self-powered system.
Professor Choi said, “I could say that this research realized a truly washable wearable electronic module in the sense that it uses daily wearable textiles instead of the plastic used in conventional wearable electronic devices. Saving energy with PSCs, it can be self-powered, using nature-friendly solar energy, and washed. I believe that it has paved the way for a ‘true-meaning wearable display’ that can be formed on textile, beyond the attachable form of wearable technology.”
This research, in collaboration with Professor Seok Ho Cho from Chonnam National University and led by Eun Gyo Jeong, was published in Energy and Environmental Science (10.1039/c8ee03271h) on January 18, 2019.
Figure 1. Schematic and photo of a washable wearing display module
Figure 2. Cover page of Energy and Environmental Science
2019.03.21 View 32370 -
Wafer-Scale Multilayer Fabrication of Silk Fibroin-Based Microelectronics
A KAIST research team developed a novel fabrication method for the multilayer processing of silk-based microelectronics. This technology for creating a biodegradable silk fibroin film allows microfabrication with polymer or metal structures manufactured from photolithography. It can be a key technology in the implementation of silk fibroin-based biodegradable electronic devices or localized drug delivery through silk fibroin patterns.
Silk fibroins are biocompatible, biodegradable, transparent, and flexible, which makes them excellent candidates for implantable biomedical devices, and they have also been used as biodegradable films and functional microstructures in biomedical applications. However, conventional microfabrication processes require strong etching solutions and solvents to modify the structure of silk fibroins.
To prevent the silk fibroin from being damaged during the process, Professor Hyunjoo J. Lee from the School of Electrical Engineering and her team came up with a novel process, named aluminum hard mask on silk fibroin (AMoS), which is capable of micropatterning multiple layers composed of both fibroin and inorganic materials, such as metal and dielectrics with high-precision microscale alignment. The AMoS process can make silk fibroin patterns on devices, or make patterns on silk fibroin thin films with other materials by using photolithography, which is a core technology in the current microfabrication process.
The team successfully cultured primary neurons on the processed silk fibroin micro-patterns, and confirmed that silk fibroin has excellent biocompatibility before and after the fabrication process and that it also can be applied to implanted biological devices.
Through this technology, the team realized the multilayer micropatterning of fibroin films on a silk fibroin substrate and fabricated a biodegradable microelectric circuit consisting of resistors and silk fibroin dielectric capacitors in a silicon wafer with large areas.
They also used this technology to position the micro-pattern of the silk fibroin thin film closer to the flexible polymer-based brain electrode, and confirmed the dye molecules mounted on the silk fibroin were transferred successfully from the micropatterns.
Professor Lee said, “This technology facilitates wafer-scale, large-area processing of sensitive materials. We expect it to be applied to a wide range of biomedical devices in the future. Using the silk fibroin with micro-patterned brain electrodes can open up many new possibilities in research on brain circuits by mounting drugs that restrict or promote brain cell activities.”
This research, in collaboration with Dr. Nakwon Choi from KIST and led by PhD candidate Geon Kook, was published in ACS AMI (10.1021/acsami.8b13170) on January 16, 2019.
Figure 1. The cover page of ACS AMI
Figure 2. Fibroin microstructures and metal patterns on a fibroin produced by using the AMoS mask.
Figure 3. Biocompatibility assessment of the AMoS Process. Top: Schematics image of a) fibroin-coated silicon b) fibroin-pattered silicon and c) gold-patterned fibroin. Bottom: Representative confocal microscopy images of live (green) and dead (red) primary cortical neurons cultured on the substrates.
2019.03.15 View 23631 -
1g-Ultrasound System for the Brain Stimulation of a Freely-moving Mouse
A KAIST research team developed a light-weight capacitive micromachined ultrasonic transducer (CMUT) and succeeded in the ultrasound brain stimulation of a freely-moving mouse. With this lightweight and compact system, researchers can conduct a versatile set of in vivo experiments.
Conventional methods for stimulating a specific brain region, such as deep brain stimulation (DBS) and optogenetics technology, are highly invasive because they have to insert probes into a target brain, which makes them difficult to use for clinical application. While transcranial magnetic stimulation (TMS) and transcranial electrical stimulation (TES) are noninvasive, they have a wide range of stimulation and problems with in-depth stimulation, which makes them problematic for target-specific treatment.
Therefore, noninvasive and focused ultrasound stimulation technology is gaining a great deal of attention as a next-generation brain stimulation alternative. Since it is delivered noninvasively, it can be applied safely in humans as well as animal experiments. Focused ultrasound stimulation is more advantageous than conventional methods in terms of providing both local and deep stimulation.
Animal behavior experiments are essential for brain stimulation research; however, ultrasonic brain stimulation technology is currently in the early stages of development. So far, only research outcomes with fixed anesthetized mice have been studied because of the heavy ultrasonic device.
Professor Hyunjoo J. Lee from the School of Electrical Engineering and her team reported a technology that can provide ultrasound stimulation to the brain of a freely-moving mouse through a microminiaturized ultrasound device.
The team studied miniaturization and ultra-lightweight CMUTs through microelectromechanical systems (MEMS) technology and designed a device suitable for behavior experiments. The device weighing less than 1g (around 0.05% of the mouse’s weight) has the center frequency, size, focal length, and ultrasonic intensity to fit a mouse’s dimensions.
To evaluate the performance of the ultrasonic device, the team stimulated the motor cortex of the mouse brain and observed the movement reaction of its forefoot. They also measured the electromyography (EMG) of the trapezius.
As a result, the team confirmed that their ultrasonic device can deliver ultrasound to a depth of 3-4mm in the mouse brain and stimulate an area of the mouse brain that represents 25% of its total size.
Based on this research, the team is investigating the effects of ultrasound on sleep by stimulating the brain of sleeping mice.
Professor Lee said, “Going beyond experimenting on fixed anesthetized mice, this research succeeded in the brain stimulation of a freely-moving mouse. We are planning to study mice with diseases, such as Parkinson’s disease, dementia, depression, and epilepsy. I believe that this basic research can contribute to treating human brain-related diseases through ultrasound brain stimulation.
This research, led by Masters candidates Hyunggug Kim and Seongyeon Kim, was published in Brain Stimulation (10.1016/j.brs.2018.11.007) on November 17, 2018.
Figure 1. The miniature transducer for the transcranial ultrasound of a freely-moving mouse
Figure 2. Its structure and simulated 2D beam profile in the axial ad radial directions
2019.03.13 View 10203 -
KAIST Develops Analog Memristive Synapses for Neuromorphic Chips
(Professor Sung-Yool Choi from the School of Electrical Engineering)
A KAIST research team developed a technology that makes a transition of the operation mode of flexible memristors to synaptic analog switching by reducing the size of the formed filament. Through this technology, memristors can extend their role to memristive synapses for neuromorphic chips, which will lead to developing soft neuromorphic intelligent systems.
Brain-inspired neuromorphic chips have been gaining a great deal of attention for reducing the power consumption and integrating data processing, compared to conventional semiconductor chips. Similarly, memristors are known to be the most suitable candidate for making a crossbar array which is the most efficient architecture for realizing hardware-based artificial neural network (ANN) inside a neuromorphic chip.
A hardware-based ANN consists of a neuron circuit and synapse elements, the connecting pieces. In the neuromorphic system, the synaptic weight, which represents the connection strength between neurons, should be stored and updated as the type of analog data at each synapse.
However, most memristors have digital characteristics suitable for nonvolatile memory. These characteristics put a limitation on the analog operation of the memristors, which makes it difficult to apply them to synaptic devices.
Professor Sung-Yool Choi from the School of Electrical Engineering and his team fabricated a flexible polymer memristor on a plastic substrate, and found that changing the size of the conductive metal filaments formed inside the device on the scale of metal atoms can make a transition of the memristor behavior from digital to analog.
Using this phenomenon, the team developed flexible memristor-based electronic synapses, which can continuously and linearly update synaptic weight, and operate under mechanical deformations such as bending.
The team confirmed that the ANN based on these memristor synapses can effectively classify person’s facial images even when they were damaged. This research demonstrated the possibility of a neuromorphic chip that can efficiently recognize faces, numbers, and objects.
Professor Choi said, “We found the principles underlying the transition from digital to analog operation of the memristors. I believe that this research paves the way for applying various memristors to either digital memory or electronic synapses, and will accelerate the development of a high-performing neuromorphic chip.”
In a joint research project with Professor Sung Gap Im (KAIST) and Professor V. P. Dravid (Northwestern University), this study was led by Dr. Byung Chul Jang (Samsung Electronics), Dr. Sungkyu Kim (Northwestern University) and Dr. Sang Yoon Yang (KAIST), and was published online in Nano Letters (10.1021/acs.nanolett.8b04023) on January 4, 2019.
Figure 1. a) Schematic illustration of a flexible pV3D3 memristor-based electronic synapse array. b) Cross-sectional TEM image of the flexible pV3D3 memristor
2019.02.28 View 10994 -
Novel Material Properties of Hybrid Perovskite Nanostructures for Next-generation Non-linear Electronic Devices
(from left: Juho Lee, Dr. Muhammad Ejaz Khan and Professor Yong-Hoon Kim)
A KAIST research team reported a novel non-linear device with the founding property coming from perovskite nanowires. They showed that hybrid perovskite-derived, inorganic-framework nanowires can acquire semi-metallicity, and proposed negative differential resistance (NDR) devices with excellent NDR characteristics that resulted from a novel quantum-hybridization NDR mechanism, implying the potential of perovskite nanowires to be realized in next-generation electronic devices.
Organic-inorganic hybrid halide perovskites have recently emerged as prominent candidates for photonic applications due to their excellent optoelectronic properties as well as their low cost and facile synthesis processes. Prominent progresses have been already made for devices including solar cells, light-emitting diodes, lasers and photodetectors.
However, research on electronic devices based on hybrid halide perovskites has not been actively pursued compared with their photonic device counterparts.
Professor Yong-Hoon Kim from the School of Electrical Engineering and his team took a closer look at low-dimensional organic-inorganic halide perovskite materials, which have enhanced quantum confinement effects, and particularly focused on the recently synthesized trimethylsulfonium (TMS) lead triiodide (CH3)3SPbI3.
Using supercomputer simulations, the team first showed that stripping the (CH3)3S or TMS organic ligands from the TMS PbI3 perovskite nanowires results in semi-metallic PbI3 columns, which contradicts the conventional assumption of the semiconducting or insulating characteristics of the inorganic perovskite framework.
Utilizing the semi-metallic PbI3 inorganic framework as the electrode, the team designed a tunneling junction device from perovskite nanowires and found that they exhibit excellent nonlinear negative differential resistance (NDR) behavior. The NDR property is a key to realizing next-generation, ultra-low-power, and multivalued non-linear devices. Furthermore, the team found that this NDR originates from a novel mechanism that involves the quantum-mechanical hybridization between channel and electrode states.
Professor Kim said, “This research demonstrates the potential of quantum mechanics-based computer simulations to lead developments in advanced nanomaterials and nanodevices. In particular, this research proposes a new direction in the development of a quantum mechanical tunneling device, which was the topic for which the Nobel Laureate in Physics in 1973 was awarded to Dr. Leo Esaki.
This research, led by Dr. Muhammad Ejaz Khan and PhD candidate Juho Lee, was published online in Advanced Functional Materials (10.1002/adfm.201807620) on January 7, 2019.
Figure. The draft version of the cover page of 'Advanced Functional Materials'
2019.02.22 View 7986 -
KAIST Develops Core Technology for Ultra-small 3D Image Sensor
(from left: Dr. Jong-Bum Yo, PhD candidate Seong-Hwan Kimand Professor Hyo-Hoon Park)
A KAIST research team developed a silicon optical phased array (OPA) chip, which can be a core component for three-dimensional image sensors. This research was co-led by PhD candidate Seong-Hwan Kim and Dr. Jong-Bum You from the National Nanofab Center (NNFC).
A 3D image sensor adds distance information to a two-dimensional image, such as a photo, to recognize it as a 3D image. It plays a vital role in various electronics including autonomous vehicles, drones, robots, and facial recognition systems, which require accurate measurement of the distance from objects.
Many automobile and drone companies are focusing on developing 3D image sensor systems, based on mechanical light detection and ranging (LiDAR) systems. However, it can only get as small as the size of a fist and has a high possibility of malfunctioning because it employs a mechanical method for laser beam-steering.
OPAs have gained a great attention as a key component to implement solid-state LiDAR because it can control the light direction electronically without moving parts. Silicon-based OPAs are small, durable, and can be mass-produced through conventional Si-CMOS processes.
However, in the development of OPAs, a big issue has been raised about how to achieve wide beam-steering in transversal and longitudinal directions. In the transversal direction, a wide beam-steering has been implemented, relatively easily, through a thermo-optic or electro-optic control of the phase shifters integrated with a 1D array. But the longitudinal beam-steering has been remaining as a technical challenge since only a narrow steering was possible with the same 1D array by changing the wavelengths of light, which is hard to implement in semiconductor processes.
If a light wavelength is changed, characteristics of element devices consisting the OPA can vary, which makes it difficult to control the light direction with reliability as well as to integrate a wavelength-tunable laser on a silicon-based chip. Therefore, it is essential to devise a new structure that can easily adjust the radiated light in both transversal and longitudinal directions.
By integrating tunable radiator, instead of tunable laser in a conventional OPA, Professor Hyo-Hoon Park from the School of Electrical Engineering and his team developed an ultra-small, low-power OPA chip that facilitates a wide 2D beam-steering with a monochromatic light source.
This OPA structure allows the minimizing of the 3D image sensors, as small as a dragonfly’s eye.
According to the team, the OPA can function as a 3D image sensor and also as a wireless transmitter sending the image data to a desired direction, enabling high-quality image data to be freely communicated between electronic devices.
Kim said, “It’s not an easy task to integrate a tunable light source in the OPA structures of previous works. We hope our research proposing a tunable radiator makes a big step towards commercializing OPAs.”
Dr. You added, “We will be able to support application researches of 3D image sensors, especially for facial recognition with smartphones and augmented reality services. We will try to prepare a processing platform in NNFC that provides core technologies of the 3D image sensor fabrication.”
This research was published in Optics Letters on January 15.
Figure 1.The manufactured OPA chip
Figure 2. Schematic feature showing an application of the OPA to a 3D image sensor
2019.02.08 View 7925 -
Enhanced Video Quality despite Poor Network Conditions
(from left: Jaehong Kim, Youngmok Jung, Hyunho Yeo, Professor Dongsu Han and Professor Jinwoo Shin)
Professor Jinwoo Shin and Professor Dongsu Han from the School of Electrical Engineering developed neural adaptive content-aware internet video delivery. This technology is a novel method that combines adaptive streaming over HTTP, the video transmission system adopted by YouTube and Netflix, with a deep learning model.
This technology is expected to create an internet environment where users can enjoy watching 4K and AV/VR videos with high-quality and high-definition (HD) videos even with weak internet connections.
Thanks to video streaming services, internet video has experienced remarkable growth; nevertheless, users often suffer from low video quality due to unfavorable network conditions. Currently, existing adaptive streaming systems adjust the quality of the video in real time, accommodating the continuously changing internet bandwidth. Various algorithms are being researched for adaptive streaming systems, but there is an inherent limitation; that is, high-quality videos cannot be streamed in poor network environments regardless of which algorithm is used.
By incorporating super-resolution in adaptive streaming, the team overcame the limit of existing content distribution networks, of which their quality relies too much on the bandwidth. In the conventional method, the server that provides the video splits a video into certain lengths of time in advance. But the novel system introduced by the team allows the downloading of neural network segments. To facilitate this method, the video server needs to provide deep neural networks for each video segment as well as sizes of Deep Neural Networks (DNN) according to the specifications of the user’s computing capacity.
The largest neural network size is two megabytes, which is considerably smaller than video. When downloading the neural network from the user’s video player, it is split into several segments. Even its partial download is sufficient for a slightly comprised super-resolution.
While playing the video, the system converts the low quality video to a high-quality version by employing super-resolution based on deep convolution neural networks (CNN). The entire process is done in real time, and users can enjoy the high-definition video.
Even with a 17% smaller bandwidth, the system can provide the Quality of Experience equivalent to the latest adaptive streaming service. At a given internet bandwidth, it can provide 43% higher average QoE than the latest service.
Using a deep learning method allows this system to achieve a higher level of compression than the existing video compression methods. Their technology was recognized as a next-generation internet video system that applies super-resolution based on a deep convolution neural network to online videos.
Professor Han said, “So far, it has only been implemented on desktops, but we will further develop applications that work in mobile devices as well. This technology has been applied to the same video transmission systems used by streaming channels such as YouTube and Netflix, and thus shows good signs for practicability.”
This research, led by Hyunho Yeo, Youngmok Jung and Jaehong Kim, was presented at the 13th UNSENIX OSDI conference on October 10 2018 and completed for filing international patent application.
For further information, please click here.
Figure 1. Image quality before (left) and after (right) the technology application
Figure 2. The technology Concept
Figure 3. A transition from low-quality to high quality video after video transmission from the video server
2019.01.22 View 7623 -
Silk Adhesive Paves the Way for Epidermal Electronics
(from left: Dr. Ji-Won Seo, Professor Hyunjoo Jenny Lee and PhD candidate, Hyojung Kim)
Producing effective epidermal electronics requires a strong, biocompatible interface between a biological surface and a sensor. Here, a KAIST team employed a calcium-modified silk fibroin as a biocompatible and strong adhesive. This technology led to the development of epidermal electronics with strong adhesion for patients who need drug injections and physiological monitoring over a long time.
Recently, biocompatible silk fibroins has been increasingly used for flexible substrates and water-soluble sacrificial layers because they allow structural modifications and are biodegradable. From previous studies, the team discovered the adhesive properties of silk fibroin via metal chelate bonding and the water-capturing of Ca ions.
Professor Hyunjoo Jenny Lee from the School of Electrical Engineering and her team explored ways to develop reusable, water-degradable, biocompatible and conductive epidermal electronics that can be attached to the human skin for long-term use. To overcome the limitations of conventional silk fibroin, the team introduced Ca ions to modify silk fibroin into a strong and biocompatible adhesive.
Calcium ions adopted in silk fibroins serve to capture water and enhance the cohesion force through metal chelation. Therefore, this endows viscoelasticity to previously a firm silk fibroin. This modified silk fibroin exhibits strong viscoelasticity and strong adhesiveness when physically attached to the human skin and various polymer substrates. Their developed silk adhesive is reusable, water-degradable, biocompatible, and conductive.
To test the effectiveness, the team employed the silk adhesive to fabricate an epidermal capacitive touch sensor that can be attached to the human skin. They verified the reusability of the sensor by performing attachment and detachment tests. They also confirmed that the physical adhesion of the Ca-modified silk facilitates its reusability and possesses high peel strength.
Furthermore, they tested the stretchability of the silk adhesive on bladder tissue. Although it is not an epidermal skin, bladder tissue is highly stretchable. Hence, it is a perfect target to measure the resistance-strain characteristic of the silk adhesive. When the bladder tissue was stretched, the resistive strain epidermal sensor corresponded to the tensile strain.
Showing high biocompatibility, the silk adhesive is suitable for interfacing with the human skin for a long period of time. Therefore, it can also be applied to a drug delivery epidermal system as well as an electrocardiogram (ECG) epidermal sensor.
Professor Lee said, “We are opening up a novel use for silk by developing reusable and biodegradable silk adhesive using biocompatible silk fibroin. This technology will contribute to the development of next-generation epidermal electronics as well as drug delivery systems.
This research, led by Dr. Ji-Won Seo and a PhD candidate, Hyojung Kim, was published in Advanced Functional Materials on September 5, 2018.
Figure 1. Schematic and photograph of a hydrogel patch adhered on the human skin through the silk adhesive
Figure 2. Cover page of Advanced Functional Materials
2018.11.21 View 7181