367 research outputs found
Writer and writing-style classification in the recognition of online handwriting
No full textOne of the problems in the automatic recognition of cursive and mixed-cursive handwriting is the large variation of handwriting styles in a population. Automatic detection of the generic handwriting style, or identification of the writer could be useful to counteract this problem. The starting point for the writing style analyses is an existing recognition system for online connected-cursive handwriting (Schomaker and Teulings, 1990; Schomaker 1993). The input to this recognizer consists of pen-tip movements produced during the writing of a single word, using equidistant sampling in time. Data are lowpass filtered and normalized on size and slant. In the segmentation stage, strokes are used, which are defined as the pen-tip trajectory between two consecutive minima in the pen-tip velocity. A neural-network technique, the Kohonen self-organizing map, is used to obtain a finite list of prototypical strokes (PS): A stroke alphabet (PSA). This stroke alphabet approximates the handwriting in the training set with a minimized rms error, and can be shown to generalize well to strokes in the handwriting of unknown writers. Thus, up to this stage of processing, the recognition system is writer independent. At the next level of processing, character classification takes place, in which sequences of stroke codes are classified as letters by a probabilistic stroke transition network. It should be noted, that at this level there is a very strong dependence on writer and writing styl
Two tree-formation methods for fast pattern search using nearest-neighbour and nearest-centroid matching
No full textThis paper describes treebased classification of character images, comparing two methods of tree formation and two methods of matching: nearest neighbor and nearest centroid. The first method, Preprocess Using Relative Distances (PURD) is a treebased reorganization of a flat list of patterns, designed to speed up nearest neighbor matching. The second method is a variant of agglomerative hierarchical clustering (HCLUS) which aims at finding a hierarchical structure of centroids in the pattern space. Results indicate that the PURD method is a very fast, effective and convenient method for the speedup of 1NN search, from which it is, however, difficult to derive usable character prototypes. HCLUS can be used to obtain very fast search with acceptable classification rate while providing character prototypes, however, at the cost of significant training efforts
Writer and writing-style classification in the recognition of online handwriting
No full textOne of the problems in the automatic recognition of cursive and mixed-cursive handwriting is the large variation of handwriting styles in a population. Automatic detection of the generic handwriting style, or identification of the writer could be useful to counteract this problem. The starting point for the writing style analyses is an existing recognition system for online connected-cursive handwriting (Schomaker and Teulings, 1990; Schomaker 1993). The input to this recognizer consists of pen-tip movements produced during the writing of a single word, using equidistant sampling in time. Data are lowpass filtered and normalized on size and slant. In the segmentation stage, strokes are used, which are defined as the pen-tip trajectory between two consecutive minima in the pen-tip velocity. A neural-network technique, the Kohonen self-organizing map, is used to obtain a finite list of prototypical strokes (PS): A stroke alphabet (PSA). This stroke alphabet approximates the handwriting in the training set with a minimized rms error, and can be shown to generalize well to strokes in the handwriting of unknown writers. Thus, up to this stage of processing, the recognition system is writer independent. At the next level of processing, character classification takes place, in which sequences of stroke codes are classified as letters by a probabilistic stroke transition network. It should be noted, that at this level there is a very strong dependence on writer and writing style<br/
Writer and writing-style classification in the recognition of online handwriting
No full textOne of the problems in the automatic recognition of cursive and mixed-cursive handwriting is the large variation of handwriting styles in a population. Automatic detection of the generic handwriting style, or identification of the writer could be useful to counteract this problem. The starting point for the writing style analyses is an existing recognition system for online connected-cursive handwriting (Schomaker and Teulings, 1990; Schomaker 1993). The input to this recognizer consists of pen-tip movements produced during the writing of a single word, using equidistant sampling in time. Data are lowpass filtered and normalized on size and slant. In the segmentation stage, strokes are used, which are defined as the pen-tip trajectory between two consecutive minima in the pen-tip velocity. A neural-network technique, the Kohonen self-organizing map, is used to obtain a finite list of prototypical strokes (PS): A stroke alphabet (PSA). This stroke alphabet approximates the handwriting in the training set with a minimized rms error, and can be shown to generalize well to strokes in the handwriting of unknown writers. Thus, up to this stage of processing, the recognition system is writer independent. At the next level of processing, character classification takes place, in which sequences of stroke codes are classified as letters by a probabilistic stroke transition network. It should be noted, that at this level there is a very strong dependence on writer and writing style<br/
Two tree-formation methods for fast pattern search using nearest-neighbour and nearest-centroid matching
No full textThis paper describes treebased classification of character images, comparing two methods of tree formation and two methods of matching: nearest neighbor and nearest centroid. The first method, Preprocess Using Relative Distances (PURD) is a treebased reorganization of a flat list of patterns, designed to speed up nearest neighbor matching. The second method is a variant of agglomerative hierarchical clustering (HCLUS) which aims at finding a hierarchical structure of centroids in the pattern space. Results indicate that the PURD method is a very fast, effective and convenient method for the speedup of 1NN search, from which it is, however, difficult to derive usable character prototypes. HCLUS can be used to obtain very fast search with acceptable classification rate while providing character prototypes, however, at the cost of significant training efforts
Un-supervised learning of prototype allographs in cursive script recognition using invariant handwriting features
No full text
- …
