Off-line Shorthand Recognition System

new technique of pattern recognition. The Pitman shorthand scripts has certain unique p hi^ work explores the possibility of fast manstructural characteristics which is the most machine communication through printed desired features for automatic recognition. shorthand scripts. A structural approach is This prompted the authors explore the used to represent and recognise printed study of pattern recognition techniques for shorthand scripts wherein the primitive recognition of printed shorthand characters . strokes and the vowels are separated in the preprocessing step . After a model-based Unlike the longhand , Pitman shorthand recognition , the primitive strokes are again notation records speech phonetically.The concatenated to identify the english word. The divides the English language into 40 efficient recognition of the shapes, position and phonems7 of 24 cons0nants912 thickness of strokes and vowels are main and d i ~ t h o n g s . Each word is features of this work. represented by separately written outline consisting of a combination of these phonetic symbols. The consonants of the word are recorded first as a connected string of strokes, hooks, loops and circles. Any vowels INTRODUCTION and dipthongs occuring in the word are written in the small marks around this applications such as verbatim reporting consonant kernel in one of the three position and office dictation, the ability to produce a relative to the associated related stroke. The simultaneous printed transcript or computer consonant kernel is also written in one of the compatible textfile of spoken language is three position relative to the writing line. often desirable. Unfortunately, typists using The thickness of a stroke distinguishes the conventional QWERTY keyboard can not between and unvoiced sound. The achieve even dictation speed of 120 wpm, consonant strokes can be halved to indicate a while typical verbatim speed is 180 wpm. following sound of t or d and it can be douAutomatic online recognition of speech offers bled to indicate a sound of rer. d e r or, d h e r . an obvious solution to this problem but the 1" standard printed script the length the recognition of connected speech of unlimited normal stroke is one half of an inch. vocabulary requires many more years of directed research. An alternative means of There has been some earlier attempt to obtaining real-time verbatim data entry is to recognise shorthand characters [ I -21 but use one of the shorthand systems [ 2 ] which t h e s t u d y " l i n e have been developed specifically for recognition. On-line recognition of shorthand verbatim recording. These systems record characters has many disadvantages. Readers speech phonetically. Handwritten shorthand are requested refer i 3 1 for d e t a i l scripts of news-reporters ( prepared at a discussion on on-line v s offline recognition remote site) require fast but off-line for handwriiten chatacters. These problems automatic conversion to equivalent English are highlighted further as the shorthand text. scripts are dependent on pen-pressure, length of the strokes etc. Our aim is to The present study is concerned with recognise complete handwritten shorthand automatic recognition of printed shorthand text and we feel that for such a purpose offand in the process exploring the line method is more practical and suitable. In the present study we concentrate on printed character as an initial step towards complete recognition.There is no earlier attempt in this direction so far.