TR
-
Mathematicians Identify a Key Source of Cell-to-Cell Variability in Cell Signaling
Systematic inferences identify a major source of heterogeneity in cell signaling dynamics
Why do genetically identical cells respond differently to the same external stimuli, such as antibiotics? This long-standing mystery has been solved by KAIST and IBS mathematicians who have developed a new framework for analyzing cell responses to some stimuli. The team found that the cell-to-cell variability in antibiotic stress response increases as the effective length of the cell signaling pathway (i.e., the number of rate-limiting steps) increases. This finding could identify more effective chemotherapies to overcome the fractional killing of cancer cells caused by cell-to-cell variability.
Cells in the human body contain signal transduction systems that respond to various external stimuli such as antibiotics and changes in osmotic pressure. When an external stimulus is detected, various biochemical reactions occur sequentially. This leads to the expression of relevant genes, allowing the cells to respond to the perturbed external environment. Furthermore, signal transduction leads to a drug response (e.g., antibiotic resistance genes are expressed when antibiotic drugs are given).
However, even when the same external stimuli are detected, the responses of individual cells are greatly heterogeneous. This leads to the emergence of persister cells that are highly resistant to drugs. To identify potential sources of this cell-to cell variability, many studies have been conducted. However, most of the intermediate signal transduction reactions are unobservable with current experimental techniques.
A group of researchers including Dae Wook Kim and Hyukpyo Hong and led by Professor Jae Kyoung Kim from the KAIST Department of Mathematical Sciences and IBS Biomedical Mathematics Group solved the mystery by exploiting queueing theory and Bayesian inference methodology. They proposed a queueing process that describes the signal transduction system in cells. Based on this, they developed Bayesian inference computational software using MBI (the Moment-based Bayesian Inference method). This enables the analysis of the signal transduction system without a direct observation of the intermediate steps. This study was published in Science Advances.
By analyzing experimental data from Escherichia coli using MBI, the research team found that cell-to-cell variability increases as the number of rate-limiting steps in the signaling pathway increases.
The rate-limiting steps denote the slowest steps (i.e., bottlenecks) in sequential biochemical reaction steps composing cell signaling pathways and thus dominates most of the signaling time. As the number of the rate-limiting steps increases, the intensity of the transduced signal becomes greatly heterogeneous even in a population of genetically identical cells. This finding is expected to provide a new paradigm for studying the heterogeneous antibiotic resistance of cells, which is a big challenge in cancer medicine.
Professor Kim said, “As a mathematician, I am excited to help advance the understanding of cell-to-cell variability in response to external stimuli. I hope this finding facilitates the development of more effective chemotherapies.”
This work was supported by the Samsung Science and Technology Foundation, the National Research Foundation of Korea, and the Institute for Basic Science.
-Publication:Dae Wook Kim, Hyukpyo Hong, and Jae Kyoung Kim (2022) “Systematic inference identifies a major source of heterogeneity in cell signaling dynamics: the rate-limiting step number,”Science Advances March 18, 2022 (DOI: 10.1126/sciadv.abl4598)
-Profile:Professor Jae Kyoung Kimhttp://mathsci.kaist.ac.kr/~jaekkim
jaekkim@kaist.ac.kr@umichkim on TwitterDepartment of Mathematical SciencesKAIST
2022.03.29 View 7219 -
Tomographic Measurement of Dielectric Tensors
Dielectric tensor tomography allows the direct measurement of the 3D dielectric tensors of optically anisotropic structures
A research team reported the direct measurement of dielectric tensors of anisotropic structures including the spatial variations of principal refractive indices and directors. The group also demonstrated quantitative tomographic measurements of various nematic liquid-crystal structures and their fast 3D nonequilibrium dynamics using a 3D label-free tomographic method. The method was described in Nature Materials.
Light-matter interactions are described by the dielectric tensor. Despite their importance in basic science and applications, it has not been possible to measure 3D dielectric tensors directly. The main challenge was due to the vectorial nature of light scattering from a 3D anisotropic structure. Previous approaches only addressed 3D anisotropic information indirectly and were limited to two-dimensional, qualitative, strict sample conditions or assumptions.
The research team developed a method enabling the tomographic reconstruction of 3D dielectric tensors without any preparation or assumptions. A sample is illuminated with a laser beam with various angles and circularly polarization states. Then, the light fields scattered from a sample are holographically measured and converted into vectorial diffraction components. Finally, by inversely solving a vectorial wave equation, the 3D dielectric tensor is reconstructed.
Professor YongKeun Park said, “There were a greater number of unknowns in direct measuring than with the conventional approach. We applied our approach to measure additional holographic images by slightly tilting the incident angle.”
He said that the slightly tilted illumination provides an additional orthogonal polarization, which makes the underdetermined problem become the determined problem. “Although scattered fields are dependent on the illumination angle, the Fourier differentiation theorem enables the extraction of the same dielectric tensor for the slightly tilted illumination,” Professor Park added.
His team’s method was validated by reconstructing well-known liquid crystal (LC) structures, including the twisted nematic, hybrid aligned nematic, radial, and bipolar configurations. Furthermore, the research team demonstrated the experimental measurements of the non-equilibrium dynamics of annihilating, nucleating, and merging LC droplets, and the LC polymer network with repeating 3D topological defects.
“This is the first experimental measurement of non-equilibrium dynamics and 3D topological defects in LC structures in a label-free manner. Our method enables the exploration of inaccessible nematic structures and interactions in non-equilibrium dynamics,” first author Dr. Seungwoo Shin explained.
-PublicationSeungwoo Shin, Jonghee Eun, Sang Seok Lee, Changjae Lee, Herve Hugonnet, Dong Ki Yoon, Shin-Hyun Kim, Jongwoo Jeong, YongKeun Park, “Tomographic Measurement ofDielectric Tensors at Optical Frequency,” Nature Materials March 02, 2022 (https://doi.org/10/1038/s41563-022-01202-8)
-ProfileProfessor YongKeun ParkBiomedical Optics Laboratory (http://bmol.kaist.ac.kr)Department of PhysicsCollege of Natural SciencesKAIST
2022.03.22 View 6115 -
Decoding Brain Signals to Control a Robotic Arm
Advanced brain-machine interface system successfully interprets arm movement directions from neural signals in the brain
Researchers have developed a mind-reading system for decoding neural signals from the brain during arm movement. The method, described in the journal Applied Soft Computing, can be used by a person to control a robotic arm through a brain-machine interface (BMI).
A BMI is a device that translates nerve signals into commands to control a machine, such as a computer or a robotic limb. There are two main techniques for monitoring neural signals in BMIs: electroencephalography (EEG) and electrocorticography (ECoG).
The EEG exhibits signals from electrodes on the surface of the scalp and is widely employed because it is non-invasive, relatively cheap, safe and easy to use. However, the EEG has low spatial resolution and detects irrelevant neural signals, which makes it difficult to interpret the intentions of individuals from the EEG.
On the other hand, the ECoG is an invasive method that involves placing electrodes directly on the surface of the cerebral cortex below the scalp. Compared with the EEG, the ECoG can monitor neural signals with much higher spatial resolution and less background noise. However, this technique has several drawbacks.
“The ECoG is primarily used to find potential sources of epileptic seizures, meaning the electrodes are placed in different locations for different patients and may not be in the optimal regions of the brain for detecting sensory and movement signals,” explained Professor Jaeseung Jeong, a brain scientist at KAIST. “This inconsistency makes it difficult to decode brain signals to predict movements.”
To overcome these problems, Professor Jeong’s team developed a new method for decoding ECoG neural signals during arm movement. The system is based on a machine-learning system for analysing and predicting neural signals called an ‘echo-state network’ and a mathematical probability model called the Gaussian distribution.
In the study, the researchers recorded ECoG signals from four individuals with epilepsy while they were performing a reach-and-grasp task. Because the ECoG electrodes were placed according to the potential sources of each patient’s epileptic seizures, only 22% to 44% of the electrodes were located in the regions of the brain responsible for controlling movement.
During the movement task, the participants were given visual cues, either by placing a real tennis ball in front of them, or via a virtual reality headset showing a clip of a human arm reaching forward in first-person view. They were asked to reach forward, grasp an object, then return their hand and release the object, while wearing motion sensors on their wrists and fingers. In a second task, they were instructed to imagine reaching forward without moving their arms.
The researchers monitored the signals from the ECoG electrodes during real and imaginary arm movements, and tested whether the new system could predict the direction of this movement from the neural signals. They found that the novel decoder successfully classified arm movements in 24 directions in three-dimensional space, both in the real and virtual tasks, and that the results were at least five times more accurate than chance. They also used a computer simulation to show that the novel ECoG decoder could control the movements of a robotic arm.
Overall, the results suggest that the new machine learning-based BCI system successfully used ECoG signals to interpret the direction of the intended movements. The next steps will be to improve the accuracy and efficiency of the decoder. In the future, it could be used in a real-time BMI device to help people with movement or sensory impairments.
This research was supported by the KAIST Global Singularity Research Program of 2021, Brain Research Program of the National Research Foundation of Korea funded by the Ministry of Science, ICT, and Future Planning, and the Basic Science Research Program through the National Research Foundation of Korea funded by the Ministry of Education.
-PublicationHoon-Hee Kim, Jaeseung Jeong, “An electrocorticographic decoder for arm movement for brain-machine interface using an echo state network and Gaussian readout,” Applied SoftComputing online December 31, 2021 (doi.org/10.1016/j.asoc.2021.108393)
-ProfileProfessor Jaeseung JeongDepartment of Bio and Brain EngineeringCollege of EngineeringKAIST
2022.03.18 View 9165 -
'Fingerprint' Machine Learning Technique Identifies Different Bacteria in Seconds
A synergistic combination of surface-enhanced Raman spectroscopy and deep learning serves as an effective platform for separation-free detection of bacteria in arbitrary media
Bacterial identification can take hours and often longer, precious time when diagnosing infections and selecting appropriate treatments. There may be a quicker, more accurate process according to researchers at KAIST. By teaching a deep learning algorithm to identify the “fingerprint” spectra of the molecular components of various bacteria, the researchers could classify various bacteria in different media with accuracies of up to 98%.
Their results were made available online on Jan. 18 in Biosensors and Bioelectronics, ahead of publication in the journal’s April issue.
Bacteria-induced illnesses, those caused by direct bacterial infection or by exposure to bacterial toxins, can induce painful symptoms and even lead to death, so the rapid detection of bacteria is crucial to prevent the intake of contaminated foods and to diagnose infections from clinical samples, such as urine. “By using surface-enhanced Raman spectroscopy (SERS) analysis boosted with a newly proposed deep learning model, we demonstrated a markedly simple, fast, and effective route to classify the signals of two common bacteria and their resident media without any separation procedures,” said Professor Sungho Jo from the School of Computing.
Raman spectroscopy sends light through a sample to see how it scatters. The results reveal structural information about the sample — the spectral fingerprint — allowing researchers to identify its molecules. The surface-enhanced version places sample cells on noble metal nanostructures that help amplify the sample’s signals.
However, it is challenging to obtain consistent and clear spectra of bacteria due to numerous overlapping peak sources, such as proteins in cell walls. “Moreover, strong signals of surrounding media are also enhanced to overwhelm target signals, requiring time-consuming and tedious bacterial separation steps,” said Professor Yeon Sik Jung from the Department of Materials Science and Engineering.
To parse through the noisy signals, the researchers implemented an artificial intelligence method called deep learning that can hierarchically extract certain features of the spectral information to classify data. They specifically designed their model, named the dual-branch wide-kernel network (DualWKNet), to efficiently learn the correlation between spectral features. Such an ability is critical for analyzing one-dimensional spectral data, according to Professor Jo.
“Despite having interfering signals or noise from the media, which make the general shapes of different bacterial spectra and their residing media signals look similar, high classification accuracies of bacterial types and their media were achieved,” Professor Jo said, explaining that DualWKNet allowed the team to identify key peaks in each class that were almost indiscernible in individual spectra, enhancing the classification accuracies. “Ultimately, with the use of DualWKNet replacing the bacteria and media separation steps, our method dramatically reduces analysis time.”
The researchers plan to use their platform to study more bacteria and media types, using the information to build a training data library of various bacterial types in additional media to reduce the collection and detection times for new samples.
“We developed a meaningful universal platform for rapid bacterial detection with the collaboration between SERS and deep learning,” Professor Jo said. “We hope to extend the use of our deep learning-based SERS analysis platform to detect numerous types of bacteria in additional media that are important for food or clinical analysis, such as blood.”
The National R&D Program, through a National Research Foundation of Korea grant funded by the Ministry of Science and ICT, supported this research.
-PublicationEojin Rho, Minjoon Kim, Seunghee H. Cho, Bongjae Choi, Hyungjoon Park, Hanhwi Jang, Yeon Sik Jung, Sungho Jo, “Separation-free bacterial identification in arbitrary media via deepneural network-based SERS analysis,” Biosensors and Bioelectronics online January 18, 2022 (doi.org/10.1016/j.bios.2022.113991)
-ProfileProfessor Yeon Sik JungDepartment of Materials Science and EngineeringKAIST
Professor Sungho JoSchool of ComputingKAIST
2022.03.04 View 19358 -
KAA Recognizes 4 Distinguished Alumni of the Year
The KAIST Alumni Association (KAA) recognized four distinguished alumni of the year during a ceremony on February 25 in Seoul. The four Distinguished Alumni Awardees are Distinguished Professor Sukbok Chang from the KAIST Department of Chemistry, Hyunshil Ahn, head of the AI Economy Institute and an editorial writer at The Korea Economic Daily, CEO Hwan-ho Sung of PSTech, and President Hark Kyu Park of Samsung Electronics.
Distinguished Professor Sukbok Chang who received his MS from the Department of Chemistry in 1985 has been a pioneer in the novel field of ‘carbon-hydrogen bond activation reactions’. He has significantly contributed to raising Korea’s international reputation in natural sciences and received the Kyungam Academic Award in 2013, the 14th Korea Science Award in 2015, the 1st Science and Technology Prize of Korea Toray in 2018, and the Best Scientist/Engineer Award Korea in 2019. Furthermore, he was named as a Highly Cited Researcher who ranked in the top 1% of citations by field and publication year in the Web of Science citation index for seven consecutive years from 2015 to 2021, demonstrating his leadership as a global scholar.
Hyunshil Ahn, a graduate of the School of Business and Technology Management with an MS in 1985 and a PhD in 1987, was appointed as the first head of the AI Economy Institute when The Korea Economic Daily was the first Korean media outlet to establish an AI economy lab. He has contributed to creating new roles for the press and media in the 4th industrial revolution, and added to the popularization of AI technology through regulation reform and consulting on industrial policies.
PSTech CEO Hwan-ho Sung is a graduate of the School of Electrical Engineering where he received an MS in 1988 and a PhD in EMBA in 2008. He has run the electronics company PSTech for over 20 years and successfully localized the production of power equipment, which previously depended on foreign technology. His development of the world’s first power equipment that can be applied to new industries including semiconductors and displays was recognized through this award.
Samsung Electronics President Hark Kyu Park graduated from the School of Business and Technology Management with an MS in 1986. He not only enhanced Korea’s national competitiveness by expanding the semiconductor industry, but also established contract-based semiconductor departments at Korean universities including KAIST, Sungkyunkwan University, Yonsei University, and Postech, and semiconductor track courses at KAIST, Sogang University, Seoul National University, and Postech to nurture professional talents. He also led the national semiconductor coexistence system by leading private sector-government-academia collaborations to strengthen competence in semiconductors, and continues to make unconditional investments in strong small businesses.
KAA President Chilhee Chung said, “Thanks to our alumni contributing at the highest levels of our society, the name of our alma mater shines brighter. As role models for our younger alumni, I hope greater honours will follow our awardees in the future.”
2022.03.03 View 6045 -
Five Projects Ranked in the Top 100 for National R&D Excellence
Five KAIST research projects were selected as the 2021 Top 100 for National R&D Excellence by the Ministry of Science and ICT and the Korea Institute of Science & Technology Evaluation and Planning.
The five projects are:-The development of E. coli that proliferates with only formic acid and carbon dioxide by Distinguished Professor Sang Yup Lee from the Department of Chemical and Biomolecular Engineering
-An original reverse aging technology that restores an old human skin cell into a younger one by Professor Kwang-Hyun Cho from the Department of Bio and Brain Engineering-The development of next-generation high-efficiency perovskite-silicon tandem solar cells by Professor Byungha Shin from the Department of Materials Science and Engineering-Research on the effects of ultrafine dust in the atmosphere has on energy consumption by Professor Jiyong Eom from the School of Business and Technology Management-Research on a molecular trigger that controls the phase transformation of bio materials by Professor Myungchul Kim from the Department of Bio and Brain Engineering
Started in 2006, an Evaluation Committee composed of experts in industries, universities, and research institutes has made the preliminary selections of the most outstanding research projects based on their significance as a scientific and technological development and their socioeconomic effects. The finalists went through an open public evaluation. The final 100 studies are from six fields: 18 from mechanics & materials, 26 from biology & marine sciences, 19 from ICT & electronics, 10 from interdisciplinary research, and nine from natural science and infrastructure.
The selected 100 studies will receive a certificate and an award plaque from the minister of MSIT as well as additional points for business and institutional evaluations according to appropriate regulations, and the selected researchers will be strongly recommended as candidates for national meritorious awards.
In particular, to help the 100 selected research projects become more accessible for the general public, their main contents will be provided in a free e-book ‘The Top 100 for National R&D Excellence of 2021’ that will be available from online booksellers.
2022.02.17 View 8193 -
President Lee Presents Plans to Nurture Next-Generation Talents
President Lee stressed that nurturing medical scientists, semiconductor R&D personnel, startup entrepreneurs, and global innovators are key missions he will continue to pursue during a news conference
KAIST President Kwang Hyung Lee said that nurturing medical scientists, semiconductor R&D personnel, startup entrepreneurs, and global innovators are key missions he will continue to pursue during an online news conference marking the 1st anniversary of him becoming the president on February 15.
He said that nurturing physician-scientists is the most critical mission for KAIST to help the nation create a new growth engine. He said KAIST will help the nation drive the bio-industry and provide medical science resources for the nation’s health sector. To this end, he said that KAIST will open its Medical Science and Technology School by 2026.
“We plan to expand the current Graduate School of Medical Science and Engineering into a new Medical Science and Technology School that will focus entirely on a condensed MD-PhD course converging the fields of AI, bio, and physics,” he said.
The school aims to foster medical scientists whose research results will eventually be commercialized. He said that the university is now discussing revisions to related laws and regulations with the government and other universities.
To supply human resources to the semiconductor industry, President Lee said the university will add a campus in Pyongtaek City that will serve as an advanced convergence research hub in the field of next generation semiconductors in collaboration with Samsung Electronics and the city of Pyongtaek. The three-stage opening plan projected the final opening of the campus by 2036. During the first stage, which will be completed by 2026, it will construct the campus infrastructure in Pyongtaek city where Samsung Semiconductors runs two massive semiconductor complexes. By 2031, it plans to launch the open research platform including a future cities research center and future vehicles research center. The campus will open the global industrial collaboration cluster hub by 2036.
In the global arena, President Lee said he is working to open the New York campus with stakeholders in the United States. He announced the plan last December that was endorsed by New York-based entrepreneur Hee-Nam Bae, the chairman of Big Continent Inc. President Lee and Chairman Lee signed an MOU for the funding to open the campus in New York.
“We are discussing how to facilitate the plan and best accommodate the interests and potential of our students. Many ideas and plans are on the table and we think it will take longer than expected to finalize the plan,” explained President Lee.
However, he added that the basic idea is to offer art tech and health technology programs as well as an AI-based finance MBA at the New York campus, in addition to it serving as the startup accelerator of KAIST.
President Lee stressed the importance of technology commercialization when successfully launching KAIST Holdings last month to help spinoffs of KAIST labs accelerate their end results. He said that KAIST Holdings will build a virtuous supporting system to commercialize the technology startups coming from KAIST.
“We plan to list at least 10 KAIST startups on the KOSDAQ and two on the NASDAQ by 2031. KAIST Holdings also aims to nurture companies valued at a total of one billion KRW and earn 100 billion KRW in technology fees by 2031.
2022.02.17 View 8047 -
Research Finds Digital Music Streaming Consumption Dropped as a Result of Covid-19 and Lockdowns
Decline in human mobility has stunning consequences for content streaming
The Covid-19 pandemic and lockdowns significantly reduced the consumption of audio music streaming in many countries as people turned to video platforms. On average, audio music consumption decreased by 12.5% after the World Health Organization’s (WHO) pandemic declaration in March 2020.
Music streaming services were an unlikely area hit hard by the Covid-19 pandemic. New research in Marketing Science found that the drop in people’s mobility during the pandemic significantly reduced the consumption of audio music streaming. Instead, people turned more to video platforms.
“On average, audio music consumption decreased by more than 12% after the World Health Organization’s (WHO) pandemic declaration on March 11, 2020. As a result, during the pandemic, Spotify lost 838 million dollars of revenue in the first three quarters of 2020,” said Jaeung Sim, a PhD candidate in management engineering at KAIST and one of the authors of the research study on this phenomenon. “Our results showed that human mobility plays a much larger role in the audio consumption of music than previously thought.”
The study, “Frontiers: Virus Shook the Streaming Star: Estimating the Covid-19 Impact on Music Consumption,” conducted by Sim and Professor Daegon Cho of KAIST, Youngdeok Hwang of City University of New York, and Rahul Telang of Carnegie Mellon University, looked at online music streaming data for top songs for two years in 60 countries, as well as Covid-19 cases, lockdown statistics, and daily mobility data, to determine the nature of the changes.
The study showed how the pandemic adversely impacted music streaming services despite the common expectation that the pandemic would universally benefit online medias platforms. This implies that the substantially changing media consumption environment can place streaming music in fiercer competition with other media forms that offer more dynamic and vivid experiences to consumers.
The researchers found that music consumption through video platforms was positively associated with the severity of Covid-19, lockdown policies, and time spent at home.
-PublicationJaeung Sim, Daegon Cho, Youngdeok Hwang, and Rahul Telang,“Frontiers: Virus Shook the Streaming Star: Estimating the Covid-19 Impact on Music Consumption,” November 30 in Marketing Science online (doi.org/10.1287/mksc.2021.1321)
-Profile Professor Daegon ChoGraduate School of Information and Media ManagementCollege of BusinessKAIST
2022.02.15 View 7405 -
Eco-Friendly Micro-Supercapacitors Using Fallen Leaves
Green micro-supercapacitors on a single leaf could easily be applied in wearable electronics, smart houses, and IoTs
A KAIST research team has developed graphene-inorganic-hybrid micro-supercapacitors made of fallen leaves using femtosecond laser direct writing. The rapid development of wearable electronics requires breakthrough innovations in flexible energy storage devices in which micro-supercapacitors have drawn a great deal of interest due to their high power density, long lifetimes, and short charging times. Recently, there has been an enormous increase in waste batteries owing to the growing demand and the shortened replacement cycle in consumer electronics. The safety and environmental issues involved in the collection, recycling, and processing of such waste batteries are creating a number of challenges.
Forests cover about 30 percent of the Earth’s surface and produce a huge amount of fallen leaves. This naturally occurring biomass comes in large quantities and is completely biodegradable, which makes it an attractive sustainable resource. Nevertheless, if the fallen leaves are left neglected instead of being used efficiently, they can contribute to fire hazards, air pollution, and global warming.
To solve both problems at once, a research team led by Professor Young-Jin Kim from the Department of Mechanical Engineering and Dr. Hana Yoon from the Korea Institute of Energy Research developed a novel technology that can create 3D porous graphene microelectrodes with high electrical conductivity by irradiating femtosecond laser pulses on the leaves in ambient air. This one-step fabrication does not require any additional materials or pre-treatment.
They showed that this technique could quickly and easily produce porous graphene electrodes at a low price, and demonstrated potential applications by fabricating graphene micro-supercapacitors to power an LED and an electronic watch. These results open up a new possibility for the mass production of flexible and green graphene-based electronic devices.
Professor Young-Jin Kim said, “Leaves create forest biomass that comes in unmanageable quantities, so using them for next-generation energy storage devices makes it possible for us to reuse waste resources, thereby establishing a virtuous cycle.”
This research was published in Advanced Functional Materials last month and was sponsored by the Ministry of Agriculture Food and Rural Affairs, the Korea Forest Service, and the Korea Institute of Energy Research.
-Publication
Truong-Son Dinh Le, Yeong A. Lee, Han Ku Nam, Kyu Yeon Jang, Dongwook Yang, Byunggi Kim, Kanghoon Yim, Seung Woo Kim, Hana Yoon, and Young-jin Kim, “Green Flexible Graphene-Inorganic-Hybrid Micro-Supercapacitors Made of Fallen Leaves Enabled by Ultrafast Laser Pulses," December 05, 2021, Advanced Functional Materials (doi.org/10.1002/adfm.202107768)
-ProfileProfessor Young-Jin KimUltra-Precision Metrology and Manufacturing (UPM2) LaboratoryDepartment of Mechanical EngineeringKAIST
2022.01.27 View 9475 -
AI Light-Field Camera Reads 3D Facial Expressions
Machine-learned, light-field camera reads facial expressions from high-contrast illumination invariant 3D facial images
A joint research team led by Professors Ki-Hun Jeong and Doheon Lee from the KAIST Department of Bio and Brain Engineering reported the development of a technique for facial expression detection by merging near-infrared light-field camera techniques with artificial intelligence (AI) technology.
Unlike a conventional camera, the light-field camera contains micro-lens arrays in front of the image sensor, which makes the camera small enough to fit into a smart phone, while allowing it to acquire the spatial and directional information of the light with a single shot. The technique has received attention as it can reconstruct images in a variety of ways including multi-views, refocusing, and 3D image acquisition, giving rise to many potential applications.
However, the optical crosstalk between shadows caused by external light sources in the environment and the micro-lens has limited existing light-field cameras from being able to provide accurate image contrast and 3D reconstruction.
The joint research team applied a vertical-cavity surface-emitting laser (VCSEL) in the near-IR range to stabilize the accuracy of 3D image reconstruction that previously depended on environmental light. When an external light source is shone on a face at 0-, 30-, and 60-degree angles, the light field camera reduces 54% of image reconstruction errors. Additionally, by inserting a light-absorbing layer for visible and near-IR wavelengths between the micro-lens arrays, the team could minimize optical crosstalk while increasing the image contrast by 2.1 times.
Through this technique, the team could overcome the limitations of existing light-field cameras and was able to develop their NIR-based light-field camera (NIR-LFC), optimized for the 3D image reconstruction of facial expressions. Using the NIR-LFC, the team acquired high-quality 3D reconstruction images of facial expressions expressing various emotions regardless of the lighting conditions of the surrounding environment.
The facial expressions in the acquired 3D images were distinguished through machine learning with an average of 85% accuracy – a statistically significant figure compared to when 2D images were used. Furthermore, by calculating the interdependency of distance information that varies with facial expression in 3D images, the team could identify the information a light-field camera utilizes to distinguish human expressions.
Professor Ki-Hun Jeong said, “The sub-miniature light-field camera developed by the research team has the potential to become the new platform to quantitatively analyze the facial expressions and emotions of humans.” To highlight the significance of this research, he added, “It could be applied in various fields including mobile healthcare, field diagnosis, social cognition, and human-machine interactions.”
This research was published in Advanced Intelligent Systems online on December 16, under the title, “Machine-Learned Light-field Camera that Reads Facial Expression from High-Contrast and Illumination Invariant 3D Facial Images.” This research was funded by the Ministry of Science and ICT and the Ministry of Trade, Industry and Energy.
-Publication“Machine-learned light-field camera that reads fascial expression from high-contrast and illumination invariant 3D facial images,” Sang-In Bae, Sangyeon Lee, Jae-Myeong Kwon, Hyun-Kyung Kim. Kyung-Won Jang, Doheon Lee, Ki-Hun Jeong, Advanced Intelligent Systems, December 16, 2021 (doi.org/10.1002/aisy.202100182)
ProfileProfessor Ki-Hun JeongBiophotonic LaboratoryDepartment of Bio and Brain EngineeringKAIST
Professor Doheon LeeDepartment of Bio and Brain EngineeringKAIST
2022.01.21 View 9580 -
Team KAIST Makes Its Presence Felt in the Self-Driving Tech Industry
Team KAIST finishes 4th at the inaugural CES Autonomous Racing Competition
Team KAIST led by Professor Hyunchul Shim and Unmanned Systems Research Group (USRG) placed fourth in an autonomous race car competition in Las Vegas last week, making its presence felt in the self-driving automotive tech industry.
Team KAIST, beat its first competitor, Auburn University, with speeds of up to 131 mph at the Autonomous Challenge at CES held at the Las Vegas Motor Speedway. However, the team failed to advance to the final round when it lost to PoliMOVE, comprised of the Polytechnic University of Milan and the University of Alabama, the final winner of the $150,000 USD race.
A total of eight teams competed in the self-driving race. The race was conducted as a single elimination tournament consisting of multiple rounds of matches. Two cars took turns playing the role of defender and attacker, and each car attempted to outpace the other until one of them was unable to complete the mission.
Each team designed the algorithm to control its racecar, the Dallara-built AV-21, which can reach a speed of up to 173 mph, and make it safely drive around the track at high speeds without crashing into the other.
The event is the CES version of the Indy Autonomous Challenge, a competition that took place for the first time in October last year to encourage university students from around the world to develop complicated software for autonomous driving and advance relevant technologies. Team KAIST placed 4th at the Indy Autonomous Challenge, which qualified it to participate in this race.
“The technical level of the CES race is much higher than last October’s and we had a very tough race. We advanced to the semifinals for two consecutive races. I think our autonomous vehicle technology is proving itself to the world,” said Professor Shim.
Professor Shim’s research group has been working on the development of autonomous aerial and ground vehicles for the past 12 years. A self-driving car developed by the lab was certified by the South Korean government to run on public roads.
The vehicle the team used cost more than 1 million USD to build. Many of the other teams had to repair their vehicle more than once due to accidents and had to spend a lot on repairs. “We are the only one who did not have any accidents, and this is a testament to our technological prowess,” said Professor Shim.
He said the financial funding to purchase pricy parts and equipment for the racecar is always a challenge given the very tight research budget and absence of corporate sponsorships.
However, Professor Shim and his research group plan to participate in the next race in September and in the 2023 CES race.
“I think we need more systemic and proactive research and support systems to earn better results but there is nothing better than the group of passionate students who are taking part in this project with us,” Shim added.
2022.01.12 View 8125 -
Team KAIST to Race at CES 2022 Autonomous Challenge
Five top university autonomous racing teams will compete in a head-to-head passing competition in Las Vegas
A self-driving racing team from the KAIST Unmanned System Research Group (USRG) advised by Professor Hyunchul Shim will compete at the Autonomous Challenge at the Consumer Electronic Show (CES) on January 7, 2022. The head-to-head, high speed autonomous racecar passing competition at the Las Vegas Motor Speedway will feature the finalists and semifinalists from the Indy Autonomous Challenge in October of this year. Team KAIST qualified as a semifinalist at the Indy Autonomous Challenge and will join four other university teams including the winner of the competition, Technische Universität München.
Team KAIST’s AV-21 vehicle is capable of driving on its own at more than 200km/h will be expected to show a speed of more than 300 km/h at the race.The participating teams are:1. KAIST2. EuroRacing : University of Modena and Reggio Emilia (Italy), University of Pisa (Italy), ETH Zürich (Switzerland), Polish Academy of Sciences (Poland) 3. MIT-PITT-RW, Massachusetts Institute of Technology, University of Pittsburgh, Rochester Institute of Technology, University of Waterloo (Canada)4.PoliMOVE – Politecnico di Milano (Italy), University of Alabama 5.TUM Autonomous Motorsport – Technische Universität München (Germany)
Professor Shim’s team is dedicated to the development and validation of cutting edge technologies for highly autonomous vehicles. In recognition of his pioneering research in unmanned system technologies, Professor Shim was honored with the Grand Prize of the Minister of Science and ICT on December 9.
“We began autonomous vehicle research in 2009 when we signed up for Hyundai Motor Company’s Autonomous Driving Challenge. For this, we developed a complete set of in-house technologies such as low-level vehicle control, perception, localization, and decision making.” In 2019, the team came in third place in the Challenge and they finally won this year.
For years, his team has participated in many unmanned systems challenges at home and abroad, gaining recognition around the world. The team won the inaugural 2016 IROS autonomous drone racing and placed second in the 2018 IROS Autonomous Drone Racing Competition. They also competed in 2017 MBZIRC, ranking fourth in Missions 2 and 3, and fifth in the Grand Challenge.
Most recently, the team won the first round of Lockheed Martin’s Alpha Pilot AI Drone Innovation Challenge. The team is now participating in the DARPA Subterranean Challenge as a member of Team CoSTAR with NASA JPL, MIT, and Caltech.
“We have accumulated plenty of first-hand experience developing autonomous vehicles with the support of domestic companies such as Hyundai Motor Company, Samsung, LG, and NAVER. In 2017, the autonomous vehicle platform “EureCar” that we developed in-house was authorized by the Korean government to lawfully conduct autonomous driving experiment on public roads,” said Professor Shim.
The team has developed various key technologies and algorithms related to unmanned systems that can be categorized into three major components: perception, planning, and control. Considering the characteristics of the algorithms that make up each module, their technology operates using a distributed computing system.
Since 2015, the team has been actively using deep learning algorithms in the form of perception subsystems. Contextual information extracted from multi-modal sensory data gathered via cameras, lidar, radar, GPS, IMU, etc. is forwarded to the planning subsystem. The planning module is responsible for the decision making and planning required for autonomous driving such as lane change determination and trajectory planning, emergency stops, and velocity command generation. The results from the planner are fed into the controller to follow the planned high-level command. The team has also developed and verified the possibility of an end-to-end deep learning based autonomous driving approach that replaces a complex system with one single AI network.
2021.12.17 View 8333