Ozcan, O.Oniz, Y.Ayyildiz, M.2024-07-182024-07-1820229781665468350https://doi.org/10.1109/HORA55278.2022.9799818https://hdl.handle.net/11411/63984th International Congress on Human-Computer Interaction, Optimization and Robotic Applications, HORA 2022 -- 9 June 2022 through 11 June 2022 -- -- 180434In this work, both simulated and experimental studies have been carried out using Spiking Neural Networks (SNNs) on the handwritten digit recognition problem. The design of SNN is performed using Spike Response Model (SRM). A gradient based algorithm is applied for the learning of SNN. For the simulations, the proposed algorithms have been applied on the MNIST data set. To provide a basis for comparison, the same studies have been also performed for an equivalent non-spiking artificial neural network (ANN) structure. Following the simulated studies, experiments utilizing a multi touch IR frame have been carried out for both network structures. Two different approaches have been adopted to evaluate the performance of the proposed network models: In the first one, the training of the network structures has been accomplished on the MNIST data set, whereas the data acquired from the experimental setup have been used in the testing phase. Next, the data obtained from the experimental setup have been employed in both training and testing of the neural networks. The results of the simulations and experimental studies reveal that the SNN outperforms the conventional ANN. © 2022 IEEE.eninfo:eu-repo/semantics/closedAccessArtificial Neural NetworksHandwritten Digit RecognitionSpiking Neural NetworksCharacter RecognitionStatistical TestsData SetGradient Based AlgorithmHandwritten Digit RecognitionMulti-TouchNetwork StructuresNeural Networks StructureNeural-NetworksPerformanceSpike Response ModelsSpiking Neural NetworkNeural NetworksConference Object2-s2.0-8513397149310.1109/HORA55278.2022.9799818N/A