Kai He
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Kai He has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Lianqing Liu
Chinese Academy of Sciences
2 shared publications - Wenxue Wang
State Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang 110016, China; Institutes for Robotics and Intelligent Manufacturing, Chinese Academy of Sciences, Shenyang 110169, China
2 shared publications Fengzhen Tang
State Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences (CAS), Shenyang, China
1 shared publication- Gang Li
Jinan Railway Transportation Group Co., Ltd. (China)
1 shared publication - Imad H. Elhajj
Department of Electrical and Computer Engineering, American University of Beirut, Beirut, Lebanon
1 shared publication - Liang Zhao
State Key Laboratory of Robotics and Intelligent Systems, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, China
1 shared publication
Latest Publications
- 12 February 2023
- Published by Springer Nature in Machine Vision and Applications
- Vol. 34 (2) , 1-13
Knowledge-based hybrid connectionist models for morphologic reasoning
- 12 February 2023
- Published by Springer Nature in Machine Vision and Applications
- Vol. 34 (2) , 1-13
Knowledge-based hybrid connectionist models for morphologic reasoning
- 27 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 21 (5) , 6497-6509
Hybrid Connectionist Symbolic Model for Morphologic Recognition by Tactile Sensing
Morphology and texture detection, which are important components of tactile sensing, augment the response of human beings to external stimuli. Similarly, tactile sensing-based information acquisition systems in robots can help enhance the interactions of robots with the surroundings. The main drawback of morphology and texture sensing methods is their inability to explain and quantify sensing information, which makes it difficult to utilize prior knowledge and necessitates a new training process to fit the new task, even if the changes between the existing and new tasks are minuscule. Another drawback is its dependence on large datasets. To solve these problems, a hybrid connectionist symbolic model (HCSM) is proposed herein that combines historic symbolic knowledge and end-to-end neural networks. The symbolic model requires a smaller dataset and possesses an improved transferability of detection. Neural networks can be easily established and exhibit better fault tolerance for non-ideal samples. The HCSM model combines these advantages. Experiments with the tactile-based morphology and texture detection demonstrated that the new method can transfer the detection ability to fit new tasks without requiring additional retraining and has a 16% higher recognition precision than a convolutional neural network, LeNet, AlexNet, VGG16, and ResNet. The HCSM method with these features can broaden the range of applications of tactile sensing.
- 27 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 21 (5) , 6497-6509
Hybrid Connectionist Symbolic Model for Morphologic Recognition by Tactile Sensing
Morphology and texture detection, which are important components of tactile sensing, augment the response of human beings to external stimuli. Similarly, tactile sensing-based information acquisition systems in robots can help enhance the interactions of robots with the surroundings. The main drawback of morphology and texture sensing methods is their inability to explain and quantify sensing information, which makes it difficult to utilize prior knowledge and necessitates a new training process to fit the new task, even if the changes between the existing and new tasks are minuscule. Another drawback is its dependence on large datasets. To solve these problems, a hybrid connectionist symbolic model (HCSM) is proposed herein that combines historic symbolic knowledge and end-to-end neural networks. The symbolic model requires a smaller dataset and possesses an improved transferability of detection. Neural networks can be easily established and exhibit better fault tolerance for non-ideal samples. The HCSM model combines these advantages. Experiments with the tactile-based morphology and texture detection demonstrated that the new method can transfer the detection ability to fit new tasks without requiring additional retraining and has a 16% higher recognition precision than a convolutional neural network, LeNet, AlexNet, VGG16, and ResNet. The HCSM method with these features can broaden the range of applications of tactile sensing.