CHARACTER RECOGNITION USING
NEURAL NETWORKS
PresentedBy:
AkashMohapatra
SouravChakraborty
RajaMummidi
LataYadav
Outline
IntroductiontoCharacterRecognition
Whyusingneuralnetworks?
ProposedModel
DataSetUsedforCharacterRecognition
CaseStudy
BackpropagationAlgorithm
Feed-ForwardNeuralNetwork
TrainingofNeuralNetwork
Results
PerformanceAnalysis
WherecanweuseCharacterRecognition?
FutureScope
CharacterRecognition:Introduction
• Convertingtheimageobtainedbyscanningatextoradocument
intomachineeditableformat.
• Difficult&challengingcomputationprocess
Whyusingneuralnetworks?
• HumanMind–Deciphercharacterseasily,accurately&Speedily
(PresenceofdenselyNeuralNetworksinhismind)
• HumanEyes-OpticalMechanism
• HumanBrain-Seesinput–Comprehendsignals–Adaptiveto
minorchangesanderrorsinvisualpatterns
• HumanVisionSystem–learnsfromexperience
Backpropagationartificialmulti-layerneuralnetwork
• InspiredbyBiologicalnervoussystem:"justlikethebrain"
• Improvequalityofrecognition(comparablyhighaccuracylevels)
• Efficientandfastinsolvingfuzzy,extremelycomplexproblems
thatdon'tyieldtotraditionalalgorithmicapproaches
ProposedModel
• RecognitionofEnglishcharactersandvariationsfromasample
setofpatternsviaANNandBackpropagationalgorithm
(ImplementationinMATLAB)
• Performanceanalysisusingvariousclassificationalgorithmslike
KNN&Decisiontree
InputMatrix
• DataSet
Training
• ArtificialNeuralNetworks
• BackpropagationAlgorithm
Output
• 26alphabets
• PerformanceAnalysis