Digital Transforms in Handwriting Recognition

Digital transforms have been intensively used in the field of pattern recognition with the advance of computers. In the field of handwriting recognition, applications of such mathematical transformations ranges from character and numeral recognition, to word recognition and signature verification. This tutorial presents the fundamentals of digital transforms and their use in handwriting recognition. The tutorial is divided into three parts. The first part deals with the fundamental concepts and an overview of digital transforms. The second part deals with the implementation of digital transforms. Specifically the Discrete Fourier Transform (DFT) is explained in detail. Some computational aspects of the DFT are evaluated and some algorithms for Fast Fourier Transform are presented. In particular the “Radix-2 FFT” algorithms are discussed in detail. In the last part of the tutorial, some applications are presented: the DFT represents one of the most powerful tools for plane curve analysis and recognition. This is useful in handwriting recognition where the major information for the description and the classification of the pattern can be found in its boundary. Furthermore, Fourier Transform allows accurate analysis of plane curves and it was recently used for an experimental observation of human behaviour in recognizing handwritten characters. The results of such experiment pose suggestive questions about human ability in handwriting recognition which are now on the frontiers of the research in this field: How does man recognize handwritten pattern? Which features of a pattern are the most important for its recognition process? What about ambiguous patterns and their properties and about human recognition of ambiguous patterns? How could be automatically treated ambiguous patterns?