Processing of digital waveforms for automatic cheque codes recognition

The Magnetic Ink Character Recognition (MICR) permits simultaneously magnetic and optical character recognition. The standard E13B, developed in 1958 by the American Bankers Associations (ABA) and currently adopted in some countries such as the USA, Canada, Australia, UK, uses a special font. The paper proposes a recognition system of digital waveforms associated to the magnetic reading of cheque codes based on the standard E13B. The system consists of three steps. The first one cuts the waveform associated to each symbol in the whole cheque code; the second performs a data reduction while the last one provides classification. Results obtained by applying two techniques for data reduction that is principal component analysis (PCA) and linear discriminant analysis (LDA) are compared. The method used for making match and recognition is based on a multiclass perceptron linear discriminator. The obtained results show that a good classification rate can be reached, by dealing also with different reader devices and a large range of cheque exemplars.