Article
Computer Science, Information Systems
Xudong Wang, Guiguang Xu, He Yan, Daiyin Zhu, Ying Wen, Zehu Luo
Summary: An approach based on time-frequency analysis and asymmetric dilated convolution coordinate attention residual networks is proposed to automatically recognize twelve kinds of low probability of intercept radar signals at low Signal-to-Noise Ratios. The approach transforms radar signals into time-frequency images using Choi-Williams distribution and applies various image processing techniques to obtain high-quality images for waveform recognition. Experimental results show that the proposed approach achieves an overall recognition accuracy of 97.94% for twelve kinds of LPI radar waveforms at -8 dB SNR.
Article
Computer Science, Artificial Intelligence
Guanhua Zhang, Minjing Yu, Yong-Jin Liu, Guozhen Zhao, Dan Zhang, Wenming Zheng
Summary: In this article, a sparse DGCNN model is proposed to improve the emotion recognition performance by imposing a sparseness constraint on the graph G. The research reveals that different brain regions may have different functions and the functional relations among electrodes are possibly highly localized and sparse. The experiments show that the sparse DGCNN model has consistently better accuracy than representative methods and has good scalability.
IEEE TRANSACTIONS ON AFFECTIVE COMPUTING
(2023)
Article
Automation & Control Systems
Zhuang Ye, Jianbo Yu
Summary: A novel deep neural network called MWMNet is proposed in this article for extracting impulses from vibration signals and performing fault diagnosis. MWMNet utilizes a smoothly embedded morphological layer to filter out noise and employs multiple branches with different scales and adaptive weighted fusion to extract impulse signals. Experimental results demonstrate that MWMNet can learn fault-related features and filter out noise, outperforming other DNN models in fault diagnosis performance.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2022)
Article
Engineering, Electrical & Electronic
Weilong Zhang, Xinghai Yang, Changli Leng, Jingjing Wang, Shiwen Mao
Summary: Correctly identifying modulation types in underwater communications is a huge challenge. This study presents a deep neural network model combining RNN and CNN for automatic feature extraction and learning, achieving higher recognition accuracy and reduced recognition time.
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS
(2022)
Article
Multidisciplinary Sciences
Sofien Gannouni, Arwa Aledaily, Kais Belwafi, Hatim Aboalsamh
Summary: Recognizing emotions using biological brain signals requires accurate signal processing and feature extraction methods. This study proposes a novel and adaptive channel selection method, along with the identification of epoch instants during emotions, to enhance the accuracy of the system. Experimental results show that the proposed method outperforms existing studies in multi-class emotion recognition with an average accuracy rate exceeding 89%. The method also shows improvement in accuracy rate when compared to existing algorithms dealing with 9 emotions, by 8%.
SCIENTIFIC REPORTS
(2021)
Article
Engineering, Electrical & Electronic
Shu Shen, Xuebin Wang, Mengshi Wu, Kang Gu, Xinrong Chen, Xinyu Geng
Summary: As a promising technology in HCI, deep learning shows good potential for gesture recognition using sEMG signals. However, existing complex network structures cannot be deployed on edge devices. Lightweight neural networks face challenges in achieving both accuracy and limited computing power. To address this, we propose a flexible and modular method based on sEMG and acceleration signals for gesture recognition. With the improved channel attention mechanism, the lightweight ICA-CNN achieves satisfactory performance in accuracy and inference speed. Experimental results demonstrate a recognition accuracy of 94.24% for 49 gestures using sEMG and acceleration signals, higher than using sEMG signals alone. Additionally, a single inference of the proposed method takes only 38.6 ms on the CPU.
IEEE SENSORS JOURNAL
(2023)
Article
Engineering, Civil
Jianping Wang, Runlong Li, Xinqi Zhang, Yuan He
Summary: This paper proposes a deep learning-based interference mitigation approach for FMCW radars, which can effectively suppress mutual interference and improve target detection performance. The method utilizes dilated convolution for network construction and a designated contrastive learning strategy for training, resulting in a higher Signal-to-Interference-plus-Noise Ratio (SINR) and target detection rate.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Engineering, Multidisciplinary
Xuezhong Wang
Summary: This paper proposes a radar electronic signal recognition algorithm based on wavelet transform and deep learning, which improves the recognition effect and robustness of electronic radar through research on signal preprocessing and feature extraction methods, as well as the design of an optimized convolution neural network.
ALEXANDRIA ENGINEERING JOURNAL
(2022)
Article
Chemistry, Multidisciplinary
Sakshi Dua, Sethuraman Sambath Kumar, Yasser Albagory, Rajakumar Ramalingam, Ankur Dumka, Rajesh Singh, Mamoon Rashid, Anita Gehlot, Sultan S. Alshamrani, Ahmed Saeed AlGhamdi
Summary: This study utilizes deep learning-based methods to model and recognize speech signals with different tones through convolutional neural networks. The experimental results demonstrate that this approach achieves better accuracy and word error rate compared to existing methods in processing continuous and extensive vocabulary speech signals.
APPLIED SCIENCES-BASEL
(2022)
Article
Geochemistry & Geophysics
Xiayuan Huang, Qiao Yang, Hong Qiao
Summary: The proposed lightweight two-stream convolutional neural network for SAR target recognition extracts multilevel features for high accuracy recognition. Experimental results demonstrate improved accuracy and significantly reduced model parameters.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2021)
Article
Computer Science, Information Systems
Jianping Zhu, Xin Lou, Wenbin Ye
Summary: Mobile-RadarNet is an efficient convolutional neural network architecture that achieves high classification accuracy while keeping computational complexity extremely low, making it suitable for deployment in mobile devices.
IEEE INTERNET OF THINGS JOURNAL
(2021)
Article
Environmental Sciences
Yifang Zhang, Qiao Chen, Pengcheng Su, Dunlong Liu, Jianzhao Cui, Jilong Chen, Jianrong Ma, Qiao Xing, Fenglin Xu, Yuanchao Fan, Fangqiang Wei
Summary: This study focuses on monitoring ice avalanches in the glaciers of Jialongcuo, Tibet using infrasound sensors due to the limitations of remote sensing satellites. The study draws conclusions on the waveform and frequency characteristics of ice avalanche events and proposes models and algorithms for signal classification and recognition. The research results provide important theoretical support for the application of infrasound-based ice avalanche monitoring technology.
Article
Geochemistry & Geophysics
Jianping Wang, Runlong Li, Yuan He, Yang Yang
Summary: In this article, a prior-guided deep learning approach is proposed for interference mitigation in FMCW radars. A complex-valued convolutional neural network is utilized, and a prior feature is exploited as a regularization term to improve performance and convergence.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Computer Science, Information Systems
Thien Huynh-The, Van-Sang Doan, Cam-Hao Hua, Quoc-Viet Pham, Toan-Van Nguyen, Dong-Seong Kim
Summary: This study proposes a convolutional neural network model LPI-Net for automatic radar waveform recognition, which utilizes deep learning methods to capture intrinsic radio characteristics. The model achieves high waveform recognition accuracy by learning representational features of time-frequency images through preceding maps collection and skip connections.
IEEE WIRELESS COMMUNICATIONS LETTERS
(2021)
Article
Engineering, Electrical & Electronic
Alexandros Ninos, Juergen Hasch, Thomas Zwick
Summary: This study introduces a human machine interaction system based on air gestures, utilizing mmWave technology and machine learning to detect moving arms up to a few meters away from the sensor. The system can classify a greater variety of dynamic gestures in real time, achieving a 94.3% accuracy on test data.
IEEE SENSORS JOURNAL
(2021)