Object and Character Recognition Using Spiking Neural Network. (2018)

• Record Type: Journal Article
• Authors: Bawane, Priyanka; Gadariye, Snehali; Chaturvedi, S.; Khurshid, A.A.
• Is Part Of: Materials today. Volume 5:Number 1(2018)Part 1
• Page Start: 360
• Page End: 366
• Publication Date: 2018
• Subjects: Artificial Neural network (ANN) -- Spiking Neural Network (SNN) -- Leaky Integrate and Fire (LIF) model -- Two level network model -- Support Vector Machine (SVM) -- Extended Histogram of Gradients (ExHoG); Materials science -- Congresses -- Periodicals; 620.1
• DOI: 10.1016/j.matpr.2017.11.093
• Languages: English
• ISSNs: 2214-7853

Abstract

Abstract: Artificial neural network have been well developed so far. First two generation of neural network, have had lot of successful application. Spiking Neural Network (SNN) often referred as third generation of neural network. Which have potential to solve problems related to biological stimulation. In today's world it is the gaining interest of active research with the areas of neural network. This paper depicts the study about classification and recognition of object and various handwritten characters using one of the popular model of SNN.Leaky integrate and fire model (LIF Model) for object recognition and two level network model for character recognition will be use. In object recognition, after applying image preprocessing operations the image is undergone through various operations for feature extraction. Here energy level of an image which is a low-level image feature and feature descriptor like Extended Histogram of Gradients (ExHoG) and Edge Detection using Extended Epanechnicov Function and Non maxima Repression are used to extract the features from an image. These feature vectors are then given to a classifier to classify an object. In character recognition, Segmentation of document imagesinto text lines and words is an important step for the document understanding however, unlike machine printed documents, the segmentation of the handwritten documents is still considered as the challenging problem due to irregular spacing between words and variation of writing style depending on the person. By developing an algorithm for characters, segmentation, feature extraction of the character and objects for classification of the given characters and object, this problem can be solved. For the evaluation of LIF and two level network model of SNN it is compared with SVM.