Context-aware Handwriting Math Expression Recognition

Now the structural recognition methods almost belong to blindly analysis methodswhich just use fixed order to put each zone recognition algorithm in the eachunrecognized zone and according to the algorithm result to judge the structural type; orsearch method of symbol recognition and structural analysis at same time which searchesin the solution space to find the minimum global cost. Because of those methods basedtraversal technique so recognition speed and location recognition accuracy are limited indifferent degrees. To solve this problem, researched an online handwrites math expressionstructural analysis method which based context-aware technology. This method accordingdifferent teaching level math expression’s probability of kinds of writing processstructural and source position symbol type to establish context-aware model. Accordingto this model, to forecast and supervise the zone in which next symbol may appear. Thismethod avoids malpractice of zone recognition of blindly analysis methods. Effectivelyimprove zone recognition accuracy and recognition speed, be beneficial for promotion theuse of indirect type electronic whiteboard system.Main job and innovation contents of this paper are below:1. Thorough analysis writing process of math expression. Divide writing process intogeneral writing process、the non-general writing process of standard type and writingprocess of special symbol system three types. The non-general writing process ofstandard type focus on representative writing process of piecewise function and writingprocess of horizontal brackets structure; Writing process of special symbol system focuson representative writing process of fraction、∑structure、∏structure、integral structureand structure.2. Detailed describe CSDTW、SVM、KNN、multi-classifier four types onlinehandwriting symbol recognition solution and select the highest accuracy of CSDTWsolution to achieve the online handwriting single stroke math symbol recognition. Use thecombination strategy to achieve the online handwriting multi-stroke math symbolrecognition.3. Put forward online handwriting math expression structural analysis method whichbased context-aware technology. Design universal structure of perceptual model ofgeneral writing process、 universal structure of perceptual model of the non-general writing process of standard type and universal structure of perceptual model of writingprocess of special symbol system. In universal structure establishes teaching probabilityand location perceived code of math expression symbol link in different location link ineach teaching level. Use context-aware mode to analyze structure. Analysis processinclude important part of zone recommendation、next zone monitor、zone confirm.4. Design and implement online handwriting math expression recognition algorithmwhich based context-aware model. Use angle between connection line of two symbolcenter point and horizontal line to judge the general structure. Use method of symbolregion and center point or method of height ratio to recognize non-general structure ofstandard type and structure of special symbol system.After the experiment the result show98.2%correct rate of zone determination; theaverage time of zone determination is about3ms. Context-aware technology makeaccuracy and speed of the online handwriting math expression recognition significantlyimproved when compared with the existing achievements.The project funded by the Hunan province key discipline construction projectsduring the12th Five Year plan.