Recognition of text recorded in Pitman shorthand language (PSL) is an interesting research problem. Automatic reading of PSL and generating equivalent English text is very challenging. The most important task involved here is the accurate recognition of Pitman stroke patterns, which constitute "text" in PSL. The paper describes automatic recognition of the strokes of the PSL at word level. A pen-down to pen-up sequence makes a stroke, which is a composition of primitives. The words are separated based on pen-down and pen-up points. The features that form a word (a stroke) are grouped first. Next, primitives and their sequence are identified and passed to a recognizer which identifies the word. A tangent-based vector through the contour of a stroke identifies the consonant primitives. Any other marks close to the stroke but not associated with the contour of a stroke represent the vowel markers.
展开▼