Conventional recognition approaches do not permit noisy patterns and significant information cannot be detected from incomplete data. This paper designs an Artificial Neural Network (ANN) that can recognize digitized-freehand characters. The characters considered include (i) letters A-Z (ii) digits 0-9 and (iii) symbols or special characters (+, -, *, /, =, (, ), ^ and %). We use the competitive-learning approach to learn and recognize the digital writings. The designed network has a Graphic User Interface (GUI) where the user draws the desired input pattern in a drawing area with the help of a mouse. The input pattern is then digitized by fitting the resulting character into a 6×8 pixel grid. The character is finally trained before recognition. Implementation shows that our system can recognize noisy patterns and incomplete inputs. While correctly written inputs could learn faster, incomplete inputs took sometime to learn. The efficiency of this design is directly proportional to the number of training sets for each pattern. The design is also adaptable to other pattern classifiers.