Pattern recognition process, computer program product and mobile terminal

Abstract

Pattern recognition process that includes, starting from an input pattern: normalization of the input pattern into a normalized pattern of predetermined size; generation of a reliable pattern from the normalized pattern by using at least one morphological operator; calculation of a distance between the reliable pattern and selected templates which are selected from a template library; classification of the reliable patterns into at least one of the classes of the selected templates by means of at least one non-parametric classification method, which uses the classes of the selected templates and the calculated distances as inputs and outputs identified classes along with confidence levels.

1 . A pattern recognition process, comprising the steps of, starting from an input pattern:
 a) normalization of the input pattern into a normalized pattern of predetermined size;   b) generation of a reliable pattern from the normalized pattern by using at least one morphological operator;   c) calculation of a distance between the reliable pattern and selected templates which are selected from a template library, wherein each template belongs to a class;   d) classification of the reliable pattern into at least one of the classes of the selected templates by means of at least one non-parametric classification method, which uses said classes of the selected templates and said calculated distances as inputs and outputs identified classes along with confidence levels.   
     
     
         2 . A pattern recognition process as in  claim 1 , wherein the normalization comprises changing the thickness of character strokes of the input pattern. 
     
     
         3 . A pattern recognition process as in  claim 2 , wherein changing the thickness comprises increasing the thickness for thin character strokes and decreasing the thickness for thick character strokes. 
     
     
         4 . A pattern recognition process as in  claim 1 , wherein one of said at least one morphological operator is erosion. 
     
     
         5 . A pattern recognition process as in  claim 4 , wherein by said erosion a first reliable pattern is generated and wherein a second reliable pattern is generated by erosion of the first reliable pattern. 
     
     
         6 . A pattern recognition process as in  claim 4 , wherein said erosion is performed by using logical operations on machine octets. 
     
     
         7 . A pattern recognition process as in  claim 1 , wherein one of said at least one morphological operator is dilation. 
     
     
         8 . A pattern recognition process as in  claim 1 , wherein said selection of templates from the template library is performed by a template matching process. 
     
     
         9 . A pattern recognition process as in  claim 1 , wherein said selection of templates from the template library is performed by means of a decision tree. 
     
     
         10 . A pattern recognition process as in  claim 9 , wherein said decision tree comprises a plurality of nodes leading to a plurality of terminal nodes, each node containing a list of reliable black pixels and reliable white pixels and each terminal node forming one of said selected templates. 
     
     
         11 . A pattern recognition process as in  claim 10 , wherein said distance calculation comprises:
 calculating a local distance at each node by comparing reliable pixels of said reliable pattern with said reliable black and white pixels in said list;   incrementing at each node a running distance by the local distance, such that the running distance at the terminal node is said distance between the reliable pattern and the selected template.   
     
     
         12 . A pattern recognition process as in  claim 11 , wherein the nodes at which said running distance exceeds a predetermined threshold are discarded. 
     
     
         13 . A pattern recognition process as in  claim 1 , wherein said selection of templates from the template library is performed by means of a decision tree followed by a template matching process. 
     
     
         14 . A pattern recognition process as in  claim 1 , wherein said at least one non-parametric classification method is chosen from the group consisting of: Parzen-window density estimation, K-nearest neighbor, probabilistic neural network, radial basis function. 
     
     
         15 . A pattern recognition process as in  claim 1 , wherein said classification comprises prioritization of a predetermined set of templates among all possible templates. 
     
     
         16 . A pattern recognition process as in  claim 15 , wherein said prioritization comprises giving templates corresponding to frequently used character fonts a higher priority. 
     
     
         17 . A computer program product directly loadable into a memory of a computer, comprising software code portions for performing the steps of, starting from an input pattern:
 a) normalization of the input pattern into a normalized pattern of predetermined size;   b) generation of a reliable pattern from the normalized pattern by using at least one morphological operator;   c) calculation of a distance between the reliable pattern and selected templates which are selected from a template library, wherein each template belongs to a class;   d) classification of the reliable pattern into at least one of the classes of the selected templates by means of at least one non-parametric classification method, which uses said classes of the selected templates and said calculated distances as inputs and outputs identified classes along with confidence levels.   
     
     
         18 . A computer program product as in  claim 17 , wherein the normalization comprises changing the thickness of character strokes of the input pattern, wherein changing the thickness comprises increasing the thickness for thin character strokes and decreasing the thickness for thick character strokes, 
     
     
         19 . A computer program product as in  claim 17 , wherein one of said at least one morphological operator is erosion by which a first reliable pattern is generated, and wherein a second reliable pattern is generated by erosion of the first reliable pattern. 
     
     
         20 . A computer program product as in  claim 19 , wherein said erosion is performed by using logical operations on machine octets. 
     
     
         21 . A computer program product as in  claim 17 , wherein said selection of templates from the template library is performed by means of a decision tree. 
     
     
         22 . A computer program product as in  claim 21 , wherein said decision tree comprises a plurality of nodes leading to a plurality of terminal nodes, each node containing a list of reliable black pixels and reliable white pixels and each terminal node forming one of said selected templates;
 wherein said distance calculation comprises:
 calculating a local distance at each node by comparing reliable pixels of said reliable pattern with said reliable black and white pixels in said list; 
 incrementing at each node a running distance by the local distance, such that the running distance at the terminal node is said distance between the reliable pattern and the selected template; 
   and wherein said selection of templates comprises discarding the nodes at which said running distance exceeds a predetermined threshold.   
     
     
         23 . A computer program product as in  claim 17 , wherein said at least one non-parametric classification method is chosen from the group consisting of: Parzen-window density estimation, K-nearest neighbor, probabilistic neural network, radial basis function. 
     
     
         24 . A computer program product as in  claim 17 , wherein said classification comprises prioritization of a predetermined set of templates among all possible templates, wherein said prioritization comprises giving templates corresponding to frequently used character fonts a higher priority. 
     
     
         25 . A computer program product according to  claim 17 , stored on a computer usable medium. 
     
     
         26 . A mobile terminal having a pattern recognition process, executable on said mobile terminal and comprising software code portions for performing the steps of, starting from an input pattern:
 a) normalization of the input pattern into a normalized pattern of predetermined size;   b) generation of a reliable pattern from the normalized pattern by using at least one morphological operator;   c) calculation of a distance between the reliable pattern and selected templates which are selected from a template library, wherein each template belongs to a class;   d) classification of the reliable pattern into at least one of the classes of the selected templates by means of at least one non-parametric classification method, which uses said classes of the selected templates and said calculated distances as inputs and outputs identified classes along with confidence levels.   
     
     
         27 . A mobile terminal according to  claim 26 , wherein the mobile terminal is a smartphone. 
     
     
         28 . A mobile terminal according to  claim 26 , wherein the mobile terminal is a tablet PC.