Pengenalan Citra Huruf Dan/ Atau Angka Menggunakan Sistem Jaringan Saraf Tiruan Model Hopfield

Abstract

Alphabetic and/ or numeric recognition is a major in computer science that can help data processing. One of technique recognition of characters (letters and/ or numbers) is an artificial neural network method. This method uses the principle that the human brain consists of neurons as the processing of input to produce output based on the existing weights. This research implements an artificial neural network system performance of the Hopfield model in image recognition of letters and/ or numbers. The image of letters and/ or numbers are normalized to a certain size and then in vectorization. So that applications can recognize the image of letters and/ or numbers then before the application must go through the process of learning to recognize all the letters and/ or numbers with a particular font type. After learning, the application also creates a file to store the Hopfield weight matrix. Weighting matrix is obtained by calculating the Hopfield neural network weight matrix of letters/ numbers by multiplying the matrix transpose of the vector bipolar with bipolar matrix vector itself. The results obtained by a square matrix, called an artificial neural network weight matrix letters/ numbers. Then specify the Hopfield neural network weights by summing the (accumulated) the entire neural network weight matrix letters/ numbers. The sum of this matrix will be obtained by a matrix reminder (memories). This matrix is used to recognize letters and/ or numbers.