Various studies related to machine learning have been performed. In this study, we focus on reinforcement learning, which is one of the methods used in machine learning. In conventional reinforcement leaning, the reward function is difficult to design, because it is complex and laborious and it requires expert knowledge. In previous studies, the robot learned from outside itself, not autonomously. To solve this problem, we propose a method of robot learning through interactions with humans using sensor input, and the reward is also generated through interactions with humans but does not require additional tasks to be performed by the human. Therefore, in this method, expert knowledge is not required, and anyone can teach the robot. Our experiment confirmed that robot learning is possible through the proposed method.
{"title":"Self-Generation of Reward by Sensor Input in Reinforcement Learning","authors":"Kaoru Nikaido, K. Kurashige","doi":"10.1109/RVSP.2013.67","DOIUrl":"https://doi.org/10.1109/RVSP.2013.67","url":null,"abstract":"Various studies related to machine learning have been performed. In this study, we focus on reinforcement learning, which is one of the methods used in machine learning. In conventional reinforcement leaning, the reward function is difficult to design, because it is complex and laborious and it requires expert knowledge. In previous studies, the robot learned from outside itself, not autonomously. To solve this problem, we propose a method of robot learning through interactions with humans using sensor input, and the reward is also generated through interactions with humans but does not require additional tasks to be performed by the human. Therefore, in this method, expert knowledge is not required, and anyone can teach the robot. Our experiment confirmed that robot learning is possible through the proposed method.","PeriodicalId":6585,"journal":{"name":"2013 Second International Conference on Robot, Vision and Signal Processing","volume":"23 1","pages":"270-273"},"PeriodicalIF":0.0,"publicationDate":"2013-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81694830","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper proposes the road detection method corresponded to multi road types with Flood Fill. Flood Fill is one of the image processing methods to partition the region of input image based on RGB color model. Road detection is useful for automatic robots because the robots work on various road surface in outdoor environment. The proposed method has two features. Firstly, the method can cancel the influence of shadow on road by using HSV color model. Secondly, the method can recognize multi road types by k-nearest neighbor algorithm. By using the proposed method, the robot can select the suitable controller for road surface or the safety route. We implement the proposed method in vehicle navigation and the availability is verified by the experimental results.
{"title":"Road Detection Method Corresponded to Multi Road Types with Flood Fill and Vehicle Control","authors":"Tomoya Fukukawa, Yu Maeda, K. Sekiyama, T. Fukuda","doi":"10.1109/RVSP.2013.68","DOIUrl":"https://doi.org/10.1109/RVSP.2013.68","url":null,"abstract":"This paper proposes the road detection method corresponded to multi road types with Flood Fill. Flood Fill is one of the image processing methods to partition the region of input image based on RGB color model. Road detection is useful for automatic robots because the robots work on various road surface in outdoor environment. The proposed method has two features. Firstly, the method can cancel the influence of shadow on road by using HSV color model. Secondly, the method can recognize multi road types by k-nearest neighbor algorithm. By using the proposed method, the robot can select the suitable controller for road surface or the safety route. We implement the proposed method in vehicle navigation and the availability is verified by the experimental results.","PeriodicalId":6585,"journal":{"name":"2013 Second International Conference on Robot, Vision and Signal Processing","volume":"23 1","pages":"274-277"},"PeriodicalIF":0.0,"publicationDate":"2013-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83577278","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Kernel learning is an important research topic in the machine learning area. Research on self-optimization learning of kernel function and its parameter has an important theoretical value for solving the kernel selection problem widely endured by kernel learning machine, and has the same important practical meaning for the improving of kernel learning systems. In this paper, we focus on two schemes: kernel optimization algorithm and procedure, the framework of kernel self-optimization learning. Finally, the proposed kernel optimization is applied into popular kernel learning methods including KPCA, KDA and KLPP. Simulation results demonstrate that the kernel self-optimization is feasible to improve various kernel-based learning methods.
{"title":"Kernel-Optimized Based Machine for Image Recognition","authors":"Yun-Heng Wang, P. Fu","doi":"10.1109/RVSP.2013.29","DOIUrl":"https://doi.org/10.1109/RVSP.2013.29","url":null,"abstract":"Kernel learning is an important research topic in the machine learning area. Research on self-optimization learning of kernel function and its parameter has an important theoretical value for solving the kernel selection problem widely endured by kernel learning machine, and has the same important practical meaning for the improving of kernel learning systems. In this paper, we focus on two schemes: kernel optimization algorithm and procedure, the framework of kernel self-optimization learning. Finally, the proposed kernel optimization is applied into popular kernel learning methods including KPCA, KDA and KLPP. Simulation results demonstrate that the kernel self-optimization is feasible to improve various kernel-based learning methods.","PeriodicalId":6585,"journal":{"name":"2013 Second International Conference on Robot, Vision and Signal Processing","volume":"129 1","pages":"98-101"},"PeriodicalIF":0.0,"publicationDate":"2013-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88642086","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper presents a neural network for online topological map construction inspired by the beta oscillations and hippocampal place cell learning. In our proposed method, nodes in the topological map represent place cells (robot location) while edges connect nodes and store robot action (i.e. orientation, direction). Our proposed method (TGARAM) comprises 2 layers: the input layer and the memory layer. The input layer collects sensory information and cluster the obtained information into a set of topological nodes incrementally. In the memory layer, the clustered information is used as a topological map where nodes are associated with actions. Then, topological nodes are clustered together into space regions to represent the environment in the memory layer. The advantages of the proposed method are that 1) it does not require high-level cognitive processes and prior knowledge which is able to work in natural environment, 2) it can process multiple sensory sources simultaneously in continuous space, and 3) it is an incremental and unsupervised learning method. Thus, topological map generated by TGARAM is utilised for path planning to constitutes a basis for robot navigation. Finally, we validate the proposed method through several experiments.
{"title":"Biologically Inspired Topological Gaussian ARAM for Robot Navigation","authors":"W. Chin, C. Loo","doi":"10.1109/RVSP.2013.66","DOIUrl":"https://doi.org/10.1109/RVSP.2013.66","url":null,"abstract":"This paper presents a neural network for online topological map construction inspired by the beta oscillations and hippocampal place cell learning. In our proposed method, nodes in the topological map represent place cells (robot location) while edges connect nodes and store robot action (i.e. orientation, direction). Our proposed method (TGARAM) comprises 2 layers: the input layer and the memory layer. The input layer collects sensory information and cluster the obtained information into a set of topological nodes incrementally. In the memory layer, the clustered information is used as a topological map where nodes are associated with actions. Then, topological nodes are clustered together into space regions to represent the environment in the memory layer. The advantages of the proposed method are that 1) it does not require high-level cognitive processes and prior knowledge which is able to work in natural environment, 2) it can process multiple sensory sources simultaneously in continuous space, and 3) it is an incremental and unsupervised learning method. Thus, topological map generated by TGARAM is utilised for path planning to constitutes a basis for robot navigation. Finally, we validate the proposed method through several experiments.","PeriodicalId":6585,"journal":{"name":"2013 Second International Conference on Robot, Vision and Signal Processing","volume":"53 1","pages":"265-269"},"PeriodicalIF":0.0,"publicationDate":"2013-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90362001","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
With the improvement in technology of thin-film-transistor liquid-crystal display (TFT-LCD), the resolution requirement of display becomes higher and higher. Super-resolution algorithms are used to enlarge original low-resolution (LR) images to meet the visual quality of the high-resolution (HR) display. In this research, an efficient super resolution algorithm is proposed. The proposed algorithm consists of two steps. First, the Lanczos interpolation is used for LR images to get the preliminary HR images. For solving the over-smoothing problems generally caused by interpolation, it needs to add texture information to refine the preliminary HR images. Subsequently, a refinement process based on simple linear regression and the self-similarity between a pair of LR and HR images is performed to provide proper information of textures. In the experimental results, the proposed algorithm not only performs well in the objective measurement such as PSNR, but also in visual qualities.
{"title":"An Efficient Super Resolution Algorithm Using Simple Linear Regression","authors":"S. Tai, Tse-Ming Kuo, Kuo-Hao Li","doi":"10.1109/RVSP.2013.71","DOIUrl":"https://doi.org/10.1109/RVSP.2013.71","url":null,"abstract":"With the improvement in technology of thin-film-transistor liquid-crystal display (TFT-LCD), the resolution requirement of display becomes higher and higher. Super-resolution algorithms are used to enlarge original low-resolution (LR) images to meet the visual quality of the high-resolution (HR) display. In this research, an efficient super resolution algorithm is proposed. The proposed algorithm consists of two steps. First, the Lanczos interpolation is used for LR images to get the preliminary HR images. For solving the over-smoothing problems generally caused by interpolation, it needs to add texture information to refine the preliminary HR images. Subsequently, a refinement process based on simple linear regression and the self-similarity between a pair of LR and HR images is performed to provide proper information of textures. In the experimental results, the proposed algorithm not only performs well in the objective measurement such as PSNR, but also in visual qualities.","PeriodicalId":6585,"journal":{"name":"2013 Second International Conference on Robot, Vision and Signal Processing","volume":"45 1","pages":"287-290"},"PeriodicalIF":0.0,"publicationDate":"2013-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76829953","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ke-Yu Lee, C. Hsiung, Keng-Chih Hsu, Chih-Feng Huang
The study aims to report the results of experimental system and curriculum design by utilizing PC, Kinect sensing device, interface circuit, embedded controller, and actuating device. Through this system and curriculum, students learned the skills of PC control programming, signal transmission, interface circuit design, and embedded controller programming, their proficiency in system integration was also developed. In order to achieve the results of Kinect sensing motion experimental system and curriculum design, literature review was conducted to understand the experimental materials. Besides, interviews with the experts of industial circles and academic community were carried out to plan the learning background and curriculum contents. In accordance with the information collected, experimental design and syllabus were outlined and further developed.
{"title":"Establishment of a Kinect Motion-Sensing System and Its Curriculum Development","authors":"Ke-Yu Lee, C. Hsiung, Keng-Chih Hsu, Chih-Feng Huang","doi":"10.1109/RVSP.2013.39","DOIUrl":"https://doi.org/10.1109/RVSP.2013.39","url":null,"abstract":"The study aims to report the results of experimental system and curriculum design by utilizing PC, Kinect sensing device, interface circuit, embedded controller, and actuating device. Through this system and curriculum, students learned the skills of PC control programming, signal transmission, interface circuit design, and embedded controller programming, their proficiency in system integration was also developed. In order to achieve the results of Kinect sensing motion experimental system and curriculum design, literature review was conducted to understand the experimental materials. Besides, interviews with the experts of industial circles and academic community were carried out to plan the learning background and curriculum contents. In accordance with the information collected, experimental design and syllabus were outlined and further developed.","PeriodicalId":6585,"journal":{"name":"2013 Second International Conference on Robot, Vision and Signal Processing","volume":"60 1","pages":"141-143"},"PeriodicalIF":0.0,"publicationDate":"2013-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83731525","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A fuzzy sliding mode controller is designed for controlling is proposed for speed tracking control of a separately excited DC motors (SEDCMs) in this paper. The method is based on Lyapounov stability, fuzzy and variable structure control theorems. First, the methodology of the fuzzy is used to estimate the unknown nonlinearities. When the system is in the sliding mode, the adaptive gain in the designed of the sliding surface is used to tackle the mismatched and unknown load torque disturbance. The result shows that the speed tracking error can be achieved in stability asymptotically.
{"title":"Design of the Fuzzy Sliding Mode Controller for DC Motor","authors":"Chih-Chin Wen, Chien-Wen Chung, Hui-Min Wang, Yaote Chang","doi":"10.1109/RVSP.2013.52","DOIUrl":"https://doi.org/10.1109/RVSP.2013.52","url":null,"abstract":"A fuzzy sliding mode controller is designed for controlling is proposed for speed tracking control of a separately excited DC motors (SEDCMs) in this paper. The method is based on Lyapounov stability, fuzzy and variable structure control theorems. First, the methodology of the fuzzy is used to estimate the unknown nonlinearities. When the system is in the sliding mode, the adaptive gain in the designed of the sliding surface is used to tackle the mismatched and unknown load torque disturbance. The result shows that the speed tracking error can be achieved in stability asymptotically.","PeriodicalId":6585,"journal":{"name":"2013 Second International Conference on Robot, Vision and Signal Processing","volume":"835 1","pages":"196-199"},"PeriodicalIF":0.0,"publicationDate":"2013-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83845101","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Exchange rate forecasting has become a popular research topic in recent years because the problems of the forecasting model selection and the improvement on forecasting accuracy are not easy to be solved. In this study, we employ a swarm intelligence method called Interactive Artificial Bee Colony (IABC) and use nine macroeconomic factors as the input for the exchange rate forecasting. The sliding window is used in the experiment for both the training and the testing. In our experiments, we use continuous previous three days data as the training set, and use the training result to forecast the fourth day's exchange rage. Moreover, we evaluate the forecasting accuracy with three criteria, namely, Mean Square Error (MSE), Mean Absolute Error (MAE), and Root Mean Square Error (RMSE). The experimental results indicate that using IABC with the macroeconomic factors is a positive and doable way for the exchange rate forecasting.
{"title":"Using Interactive Artificial Bee Colony to Forecast Exchange Rate","authors":"Jui-Fang Chang, Chun-Tsung Hsiao, Pei-wei Tsai","doi":"10.1109/RVSP.2013.37","DOIUrl":"https://doi.org/10.1109/RVSP.2013.37","url":null,"abstract":"Exchange rate forecasting has become a popular research topic in recent years because the problems of the forecasting model selection and the improvement on forecasting accuracy are not easy to be solved. In this study, we employ a swarm intelligence method called Interactive Artificial Bee Colony (IABC) and use nine macroeconomic factors as the input for the exchange rate forecasting. The sliding window is used in the experiment for both the training and the testing. In our experiments, we use continuous previous three days data as the training set, and use the training result to forecast the fourth day's exchange rage. Moreover, we evaluate the forecasting accuracy with three criteria, namely, Mean Square Error (MSE), Mean Absolute Error (MAE), and Root Mean Square Error (RMSE). The experimental results indicate that using IABC with the macroeconomic factors is a positive and doable way for the exchange rate forecasting.","PeriodicalId":6585,"journal":{"name":"2013 Second International Conference on Robot, Vision and Signal Processing","volume":"31 1","pages":"133-136"},"PeriodicalIF":0.0,"publicationDate":"2013-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81732657","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this paper, we propose a new method for cartoon image retrieval based on the local invariant shape feature, named Scalable Shape Context. The proposed feature uses the Harris-Laplace corner to localize the key points and corresponding scale in the cartoon image. Then, we use Shape Context to describe the local shape. The feature point matching is achieved by a weighted bipartite graph matching algorithm and the similarity between the query and the indexing image is presented by the match cost. The experimental results show that our method is more efficient than Shape Context and SIFT for the cartoon image retrieval.
{"title":"Local Invariant Shape Feature for Cartoon Image Retrieval","authors":"Tiejun Zhang, Q. Han, Handan Hou, X. Niu","doi":"10.1109/RVSP.2013.31","DOIUrl":"https://doi.org/10.1109/RVSP.2013.31","url":null,"abstract":"In this paper, we propose a new method for cartoon image retrieval based on the local invariant shape feature, named Scalable Shape Context. The proposed feature uses the Harris-Laplace corner to localize the key points and corresponding scale in the cartoon image. Then, we use Shape Context to describe the local shape. The feature point matching is achieved by a weighted bipartite graph matching algorithm and the similarity between the query and the indexing image is presented by the match cost. The experimental results show that our method is more efficient than Shape Context and SIFT for the cartoon image retrieval.","PeriodicalId":6585,"journal":{"name":"2013 Second International Conference on Robot, Vision and Signal Processing","volume":"48 4 1","pages":"107-110"},"PeriodicalIF":0.0,"publicationDate":"2013-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83201989","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Zizhu Fan, Ming Ni, Meibo Sheng, Zejiu Wu, Baogen Xu
In machine learning and pattern recognition, principal component analysis (PCA) is a very popular feature extraction and dimensionality reduction method for improving recognition performance or computational effiency. It has been widely used in numerous applications, especially in face recognition. Researches often use PCA integrating the nearest neighbor classifier (NNC) based on Euclidean distance (ED) to classify face images. We refer to this method as PCA+ED. However, we have observed that PCA can not significantly improve the recognition performance of NNC based on Euclidean distance through many experiments. The main reason is that PCA can not significantly change the Euclidean distance between samples when many components are used in classification. In order to improve the classification performance in face recognition, we use another distance measure, i.e., Mahalanobis distance (MD), in NNC after performing PCA in this paper. This approach is referred to as PCA+MD. Several experiments show that PCA+MD can significantly improve the classification performance in face recognition.
{"title":"Principal Component Analysis Integrating Mahalanobis Distance for Face Recognition","authors":"Zizhu Fan, Ming Ni, Meibo Sheng, Zejiu Wu, Baogen Xu","doi":"10.1109/RVSP.2013.27","DOIUrl":"https://doi.org/10.1109/RVSP.2013.27","url":null,"abstract":"In machine learning and pattern recognition, principal component analysis (PCA) is a very popular feature extraction and dimensionality reduction method for improving recognition performance or computational effiency. It has been widely used in numerous applications, especially in face recognition. Researches often use PCA integrating the nearest neighbor classifier (NNC) based on Euclidean distance (ED) to classify face images. We refer to this method as PCA+ED. However, we have observed that PCA can not significantly improve the recognition performance of NNC based on Euclidean distance through many experiments. The main reason is that PCA can not significantly change the Euclidean distance between samples when many components are used in classification. In order to improve the classification performance in face recognition, we use another distance measure, i.e., Mahalanobis distance (MD), in NNC after performing PCA in this paper. This approach is referred to as PCA+MD. Several experiments show that PCA+MD can significantly improve the classification performance in face recognition.","PeriodicalId":6585,"journal":{"name":"2013 Second International Conference on Robot, Vision and Signal Processing","volume":"7 1","pages":"89-92"},"PeriodicalIF":0.0,"publicationDate":"2013-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89013145","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：