Recognition of English Capital Alphabet in Air writing Using Convolutional Neural Network and Intel RealSense D435 Depth Camera
Marie Claire C. Escopete, Christian G. Laluon, Ezekiel M. Llarenas, Pars M. Reyes, Roselito E. Tolentino · 2021 2nd Global Conference for Advancement in Technology (GCAT) · 2021
The study focused on the recognition of the English Capital Alphabet written in free air using Convolutional Neural Network (CNN). The previous study conducted by Balesteros et al. (2018) used a slope orientation sequence matching algorithm. Wherein, the slope defined by the plot of continuous stream of points is used to represent Alphanumeric Characters. Also, in this study the proponents achieved the system’s reliability in recognizing alphanumeric characters in average is 97%. However, some alphanumeric character slope orientation sequence is nearly identical. Letters U and V, as well as letters D and P, have identical slope orientation sequences resulting in misclassification. In order to solve this problem and improve the accuracy of recognizing air-drawn English Capital Alphabet, CNN, a deep learning algorithm is used. CNN is primarily used to search for patterns in an image. As it digs deeper, it recognizes the correct trait on its own until it gets a high-level representation of the predicted outcome. By using CNN, the system’s reliability in recognizing English Capital Alphabet characters on average is 99.8%.