School of Automation, Beijing University of Posts and Telecommunications,Beijing 100876,School of Automation, Beijing University of Posts and Telecommunications,Beijing 100876
In this paper, we have exploited an original principle of the incremental online sketchy shapes recognition. The recognition process is based on two aspects: the dynamic user modeling and relevance feedback. Firstly, we adopt an dynamic user modeling way to build user models for each specific user in an incremental decision tree and recognize dynamically the ‘possible shapes’ by means of fuzzy matching based on the visiting frequency of each recorded shape of user models. Secondly, we make attempt to bring relevance feedback method into the incremental sketchy recognition to to capture users’ input intends and refine the recognition results incrementally. Experiments prove the proposed method both effective and efficient.
Nanjing University,Nanjing University,Nanjing University
In this paper, we present a multi-strokes sketchy graphics recognition method, which make user of temporal information of sketching. The main idea includes two aspects: temporal-based stroke segmentation and temporal-based user modeling. The former means that the input stroke is emendated into primitive shapes such as lines, arcs, circles and ellipses, etc., based on pen speed and stroke curvature. The later indicates that a users’ sketching model are constructed and adjusted dynamically based on temporal order of primitive shapes. The experiments have proved the high efficiency.
Nanjing University,Nanjing Univeristy
常州大学 信息科学与工程学院，江苏 常州 213164,常州大学 信息科学与工程学院，江苏 常州 213164,常州大学 信息科学与工程学院，江苏 常州 213164
Adaptation is a critical problem in the design of user-centered human-computer interaction systems. In this paper, an SVM-based incremental learning algorithm is presented to solve this problem for sketch recognition, the goal of which is to achieve adaptive sketch recognition. Our algorithm utilizes only the support vectors instead of all the historical samples, and selects some important samples from all newly added samples as training data. The importance of a sample is measured according to its distance to the hyper-plane of the SVM classifier. Theoretical analysis, experimentation, and evaluation of our algorithm in our on-line graphics recognition system are presented to show the effectiveness of this algorithm. According to our experiments, this algorithm can reduce both the training time and the required storage space for the training dataset to a large extent with very little loss of precision.
Nanjing University,Nanjing University