Article
Computer Science, Artificial Intelligence
Akira Furui, Takuya Igaue, Toshio Tsuji
Summary: This study proposes an EMG pattern classification method that incorporates a scale mixture-based generative model and trains the model using variational Bayesian learning. An information-based method is introduced to optimize the hyperparameters of the proposed method. Experimental results demonstrate the superiority of the proposed method on public datasets and validate its effectiveness.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Engineering, Mechanical
Haiying Liang, Huanhuan Lu, Kunpeng Feng, Yang Liu, Jintao Li, Lei Meng
Summary: The study introduces a new method called NWKL based on improved NOFRFs utilizing KL divergence to detect rub-impact faults in rotor systems. An index called KLRm is proposed to evaluate the severity of rub-impact faults, demonstrating high sensitivity and feature dynamic stability. The methods show significant importance and potential application values in detecting and evaluating rotor rub-impact.
NONLINEAR DYNAMICS
(2021)
Article
Geochemistry & Geophysics
Naihao Liu, Shengtao Wei, Rongchang Liu, Yang Yang, Nan Zhang, Jinghuai Gao
Summary: In this study, an unscaled generalized S-transform (UGST) and a Q estimation method based on the weighted Kullback-Leibler divergence are proposed, and an estimation workflow is built. Compared with contrastive methods, the proposed workflow achieves more accurate estimation results and demonstrates better noise immunity.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Mengxi Liu, Pengyuan Zhang, Qian Shi, Mengwei Liu
Summary: Land cover classification is crucial for land resource monitoring and planning. Deep learning-based methods have become dominant for precise land cover mapping, but domain shift between different images hinders their large-scale application. In this study, an adversarial domain adaptation framework with Kullback-Leibler constraint is proposed to improve remote sensing land cover classification performance.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Shu Yin, Peican Zhu, Xinyu Wu, Jiajin Huang, Xianghua Li, Zhen Wang, Chao Gao
Summary: Text classification is a crucial task in various text-related downstream assignments. Recent graph-based methods, which treat text as a set of co-occurring words instead of a sequential structure, have shown excellent results in this task. However, existing corpus-level graph models struggle to incorporate local semantic information and classify new texts. To address these challenges, we propose a Global-Local Text Classification (GLTC) model that utilizes KL constraints for inductive learning. The model consists of a global structural feature extractor, a local semantic feature extractor, and a KL divergence regularization term. Experimental results demonstrate that GLTC outperforms baseline methods in terms of accuracy.
NEURAL COMPUTING & APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Son T. Duong, Ha H. Nguyen, Ebrahim Bedeer
Summary: This letter investigates non-coherent detection of SIMO systems over block Rayleigh fading channels. A multiple-symbol constellation optimization problem is formulated using the Kullback-Leibler divergence as the design criterion, which has high computational complexity. By decoupling the problem into a unitary constellation design and a multi-level design, the proposed multi-level design achieves low complexity in both construction and detection. Simulation results demonstrate that our multi-level design outperforms traditional pilot-based schemes and other existing low-complexity multi-level designs.
IEEE WIRELESS COMMUNICATIONS LETTERS
(2023)
Article
Computer Science, Artificial Intelligence
Mundher Mohammed Taresh, Ningbo Zhu, Talal Ahmed Ali Ali, Mohammed Alghaili, Asaad Shakir Hameed, Modhi Lafta Mutar
Summary: The emergence of the COVID-19 pandemic led to worldwide chaos, but the development and distribution of vaccines brought relief before a new wave hit. Early detection of infected individuals is now a top priority. This paper proposes a method using chest x-ray images from the COVIDx dataset for detection, achieving impressive results with the KL-MOB convolution neural network architecture.
PEERJ COMPUTER SCIENCE
(2021)
Article
Engineering, Electrical & Electronic
Hyeyun Lee, Soyoung Lee, Jaeseong Kim, Heesoo Jung, Kyung Jae Yoon, Srinivas Gandla, Hogun Park, Sunkook Kim
Summary: With the help of AI-based algorithms, the accuracy of gesture recognition using sEMG signals has increased. An array of bipolar stretchable sEMG electrodes, combined with a self-attention-based graph neural network, is developed to achieve highly accurate gesture recognition. The system can differentiate static and dynamic gestures with about 97% accuracy using a single trial per gesture. The array also has skin-like attributes and can provide stable EMG signals even after long-term testing and multiple reuses.
NPJ FLEXIBLE ELECTRONICS
(2023)
Article
Engineering, Geological
Q. J. Pan, X. Z. Li, S. Y. Wang, K. K. Phoon
Summary: This study performs a statistically rigorous model comparison and a probabilistic assessment of information gain for different types of monitoring data in tunneling-induced ground deformation analysis. The results indicate that the Loganathan-Poulos model is the most suitable for predicting tunneling-induced ground deformations. The analysis also reveals that measured ground vertical deformations are more informative than measured ground horizontal deformations.
Article
Engineering, Electrical & Electronic
Faraz Amjad, Santhosh Kumar Varanasi, Biao Huang
Summary: This paper presents a computer vision method based on CNNs and KF for detecting interfaces in PSV, with experimental results demonstrating accuracy and robustness to different abnormalities in the process.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Engineering, Biomedical
Jianan Li, Ziling Zhu, William J. Boyd, Carlos Martinez-Luna, Chenyun Dai, Haopeng Wang, He Wang, Xinming Huang, Todd R. Farrell, Edward A. Clancy
Summary: Most transradial prosthesis users have limited function with conventional Sequential myoelectric control, as they can only control one degree of freedom at a time. However, our regression-based EMG control method allows simultaneous and proportional control of two degrees of freedom in a virtual task. Using a short calibration period and automated electrode site selection, we achieved better target matching performance than Sequential control.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2023)
Article
Chemistry, Multidisciplinary
Ge Zhang, Qiong Yang, Guotong Li, Jiaxing Leng, Long Wang
Summary: This paper studies the detection of incipient faults in satellites using projection vector and KL divergence, showing that the PVs obtained by PCA are not necessarily optimal for detection. A numerical example and a real satellite fault case were used to evaluate the method, resulting in higher fault detection rates compared to conventional methods.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Information Systems
Lee-Kien Foo, Sook-Ling Chua, Neveen Ibrahim
Summary: The naive Bayes classifier is a simple yet effective method for data mining classification. However, the assumption of attribute independence may not hold in real-world applications. To address this, researchers proposed a method to incorporate attribute weights into naive Bayes, which resulted in improved classification performance in terms of accuracy and F1 score.
CMC-COMPUTERS MATERIALS & CONTINUA
(2022)
Article
Automation & Control Systems
Yuexi Wang, Tetsuya Kaji, Veronika Rockova
Summary: Approximate Bayesian Computation (ABC) is a method that enables statistical inference in simulator-based models with difficult likelihood calculations but easy simulation. This study constructs a kernel-type approximation of the posterior distribution in ABC by comparing summary statistics of real and simulated data, and uses contrastive learning to directly compare empirical distributions. The research demonstrates the usefulness of this approach in simulated examples and real data analysis in the context of stock volatility estimation.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Chemistry, Multidisciplinary
Wafa Shafqat, Yung-Cheol Byun
Summary: This paper proposed a Graph Convolutional Neural Network (GCNN)-based approach for online product recommendation, facing challenges in handling computational complexities and training large datasets.
APPLIED SCIENCES-BASEL
(2021)
Article
Clinical Neurology
Koji Shimonaga, Seiji Hama, Toshio Tsuji, Kazumasa Yoshimura, Shinya Nishino, Akiko Yanagawa, Zu Soh, Toshinori Matsushige, Tatsuya Mizoue, Keiichi Onoda, Hidehisa Yamashita, Shigeto Yamawaki, Kaoru Kurisu
Summary: This study used MRI and cognitive function tests to explore specific brain structure abnormalities associated with post-stroke cognitive dysfunction related to driving ability, highlighting the crucial role of the right hemisphere in maintaining cognitive function and car driving ability.
NEUROSURGICAL REVIEW
(2021)
Article
Biology
Michiyo Suzuki, Zu Soh, Hiroki Yamashita, Toshio Tsuji, Tomoo Funayama
Article
Computer Science, Interdisciplinary Applications
Chiara Tacchino, Martina Impagliazzo, Erika Maggi, Marta Bertamino, Isa Blanchi, Francesca Campone, Paola Durand, Marco Fato, Psiche Giannoni, Riccardo Iandolo, Massimiliano Izzo, Pietro Morasso, Paolo Moretti, Luca Ramenghi, Keisuke Shima, Koji Shimatani, Toshio Tsuji, Sara Uccella, Nicolo Zanardi, Maura Casadio
Summary: A simple video analysis system, MIMAS, was developed for analyzing spontaneous movements of preterm babies, showing sensitivity to newborns' maturation level and birth conditions, as well as capable of detecting abnormal preterm babies.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2021)
Article
Engineering, Biomedical
Akira Furui, Ryota Onishi, Akihito Takeuchi, Tomoyuki Akiyama, Toshio Tsuji
Summary: This article proposed a stochastic EEG model based on multivariate scale mixture distribution to represent non-Gaussian characteristics associated with epileptic seizures. Experimental results demonstrated the validity and applicability of the model in epileptic seizure detection with high accuracy.
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING
(2021)
Article
Multidisciplinary Sciences
Keisuke Shima, Koji Shimatani, Mami Sakata
Summary: This study introduces a new approach of virtual light-touch contact and proposes a wearable light-touch system that generates vibrotactile fingertip feedback to mitigate postural sway. Validation experiments suggest that this method supports human balance on a comparable level to light-touch effect.
SCIENTIFIC REPORTS
(2021)
Article
Multidisciplinary Sciences
Kazuma Sakamoto, Zu Soh, Michiyo Suzuki, Yuichi Iino, Toshio Tsuji
Summary: In short, the study investigates how the neural network of C. elegans can generate oscillations for both forward and backward movement and analyzes the distribution of connection weights through machine learning. The research confirms that the motor neuron circuit can generate oscillations for forward and backward movement, with connection weights following a Boltzmann-type distribution.
SCIENTIFIC REPORTS
(2021)
Article
Computer Science, Artificial Intelligence
Akira Furui, Takuya Igaue, Toshio Tsuji
Summary: This study proposes an EMG pattern classification method that incorporates a scale mixture-based generative model and trains the model using variational Bayesian learning. An information-based method is introduced to optimize the hyperparameters of the proposed method. Experimental results demonstrate the superiority of the proposed method on public datasets and validate its effectiveness.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Multidisciplinary Sciences
Prasetia Utama Putra, Keisuke Shima, Koji Shimatani
Summary: This study introduces a DNNs model that utilizes multiple cameras, trained using transfer learning and shared-weight techniques. Experimental results demonstrate that the model performs well on challenging datasets, and competitive recognition rates are observed in online classification.
Article
Multidisciplinary Sciences
Prasetia Utama Putra, Keisuke Shima, Sergio A. Alvarez, Koji Shimatani
Summary: Previous studies have shown that ASD children exhibit more unstable gaze modulation during a Go/No-Go task compared to typical children. By tracking children's gaze and using an algorithm, it is possible to accurately differentiate between ASD and typical children.
SCIENTIFIC REPORTS
(2021)
Article
Robotics
Hayato Mikami, Mami Sakata, Keisuke Shima, Tomoyuki Nagara, Makiko Yamashita, Koji Shimatani
Summary: This paper developed a simple and quantitative evaluation system for a single-leg standing test. The proposed system utilizes a camera with a depth sensor and a force plate to measure the movement and represents it as a state transition model. The experiment results show that the system can objectively evaluate the movement of each state, even in worksites without skilled evaluators.
Article
Engineering, Biomedical
Ziqiang Xu, Toshiki Sakagawa, Akira Furui, Shumma Jomyo, Masanori Morita, Masamichi Ando, Toshio Tsuji
Summary: This study reports a method to estimate rapid changes in peripheral arterial stiffness induced by sympathetic nervous system activity (SNSA) using local pulse wave velocity (LPWV), and further quantifies SNSA based on the estimated stiffness. The validity of LPWV in the quantitative evaluation of SNSA is demonstrated, and blood pressure correction can be optional depending on application scenarios.
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING
(2022)
Article
Neurosciences
Shun Sawai, Shin Murata, Shoya Fujikawa, Ryosuke Yamamoto, Keisuke Shima, Hideki Nakano
Summary: This study aimed to examine the effect of applying tDCS directly before MI with NFB. The results showed that m-ERD significantly increased in the NFB + tDCS group compared to the NFB group, and MI vividness significantly improved before and after training. This indicates that the combination of tDCS and NFB is more effective in improving MI abilities.
FRONTIERS IN NEUROSCIENCE
(2023)
Proceedings Paper
Engineering, Biomedical
Ryotaro Watanabe, Keisuke Shima, Taiki Horiuchi, Takeshi Shimizu, Takayuki Mukaeda, Koji Shimatani
Summary: This paper proposes an evaluation/treatment support system that enables automatic determination of wound evaluation indices from RGB-depth images and fully convolutional networks. Experimental results show that the technique reduces errors and achieves higher levels of tissue classification compared to previous approaches.
2021 43RD ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE & BIOLOGY SOCIETY (EMBC)
(2021)
Proceedings Paper
Automation & Control Systems
Prasetia Utama Putra, Keisuke Shima, Sayaka Hotchi, Koji Shimatani
Summary: This study introduces a markerless behavior evaluation system that utilizes multiple RGB cameras and Kinect V2 sensors to assist clinicians in identifying disorder symptoms in children. By tracking children's and toys' positions and recording their activities, the system can estimate children's behavior and extract features for modeling. Experimental results indicate that the frequency of changing activities and playing alone are more informative in distinguishing ASD children from typical ones.
2021 21ST INTERNATIONAL CONFERENCE ON CONTROL, AUTOMATION AND SYSTEMS (ICCAS 2021)
(2021)
Article
Computer Science, Information Systems
Hideaki Hayashi, Taro Shibanoki, Toshio Tsuji
Summary: This paper proposes a neural network based on the Johnson S-U translation system that can effectively represent data with skewness and kurtosis distributions. By transforming the discriminative model into a linear combination and incorporating it into the neural network structure, the convergence of network learning is theoretically guaranteed. Experimental results show that the proposed network achieved high classification performance without the need for hyperparameter optimization on artificially generated and real biological data.