Font Size: a A A

Research On Freehand Symbol Recognition And System Design

Posted on:2014-06-06Degree:MasterType:Thesis
Country:ChinaCandidate:Z F ZhouFull Text:PDF
GTID:2268330422963509Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Freehand symbol recognition technology is an important problem in man-machineinteraction research. It has been greatly developed in recent years. Its application areasrelated to the civil and military direction such as digital formula recognition in the processof e-learning, the identification of electronic components, road sign recognition andidentification of military label have a wide range of coverage. All these applicationsrequire high recognition accuracy and fast on-line reaction, so the symbol sketchrecognition algorithm must have a high recognition rate and the real-time nature. But thecurrent sketch recognition technology cannot meet these two requirements, especially forthe high side of the complex symbols as well as the number of symbols. To improve therecognition rate is really difficult. For some problems that exist in the traditional symbolsketch recognition, I have made some innovative work in the following aspects.To extract the ratio features of the basic sketch based on the idea of fuzzy logic, andtrain classifiers offline using support vector machines. Identify the basic symbols sketchbased on ratio characteristics and get better recognition rate. Find the idea of the basicsymbols sketches drawn separately, not only to bring convenience for user, but also tofurther improve the recognition rate of the basic symbols sketch.Introduce a fast indexing mechanism for the combination symbols sketch recognition.On the basis of the basic symbols sketch recognition, we establish a combination symbolssketch primitives index table to quickly locate relevant combinations of symbols sketchbased on the combination symbols of the input sketch indexed array, this can greatlynarrow the identification range, then introduce the rule base and global features ofmachine learning methods to have a secondary identification within this narrow range.Achieve satisfactory recognition results.Complete the symbol sketch online identification system. It can provide onlineinteraction mechanisms for users. The system can also give sort of recognition results inreal time.
Keywords/Search Tags:freehand symbol recognition, fuzzy logic, ratio features, draw separately, rulebase
PDF Full Text Request
Related items