Human Feature Perception as a Complementary Method for Digit Recognition

E. Vzquez-Fernández, H. Gonzlez-Jorge, . Dacal-Nieto, F. Martn, and A. Formella (Spain)


Machine vision, Digit recognition, Classifier fusion,Digital instrumentation.


In this paper, we propose a novel approach for the recognition of multiple format digits that is based on human perception. To show the benefits of using this system, we have tested its performance in a real working application and we show the obtained results. We have developed an automatic calibration system for digital measurement instrumentation using our application. After having tested different methods for digit recognition, we designed a fusion scheme to benefit from the different capabilities of two recognizers. We combined the use of feature extraction and a classical distance classifier with a new recognizer based on simple feature perception. This last one tries to emulate the way people select features to recognize characters regardless of their font type. The way we combine both methods makes the fusion a stronger system: we improved the recognition rate for entire display strings from 79.5% to 99.3%.

Important Links:

Go Back