Handwritten digit recognition using elastic matching

Handwritten digits recognition is a well-researched sub area within the field of handwritten writing recognition, which is concerned with learning models to distinguish presegmented handwritten digits. However, handwritten digit recognition is still a difficult task because of the high variability in the digit shapes written by individuals. So, we proposed a new elastic image matching (EM) technique for recognition of offline isolated English handwritten digits. Elastic matching can be performed by matching unknown against template or a sequence of template. Our work is divided into two main steps: feature extraction and recognition. The first step involves the pre-processing of the image to reduce some undesirable variability that only contributes to complicate the recognition process. Operations like filtering, normalization, segmentation etc. are carried out at this stage. Feature extraction is done using elastic matching is probably the most important factor in achieving high recognition performance in digit recognition systems. The second step is the digit recognition. Here we used template matching which is based on a similarity measure (e.g.: Euclidean, Mahalanobis similarity measures etc.). A prototype matching is done for recognition.