Dysgraphia Identification from Handwriting with Support Vector Machine Method

Abstract

Dysgraphia, a handwriting disorder in which a person has difficulty in writing at any level such as slow writing or unreadable letter. Many research has done to study the characteristics and to diagnose it for early prevention in children. In this study, we try to identify dysgraphia among children and divide it into 4 class, normal, light, moderate, and severe. Therefore an android application with embedding a handwriting recognition tool was created to collect the data from elementary school students that have dysgraphia and those who don’t. We use Support Vector Machine in classifying the data to identify dysgraphia because SVM has the ability to learn well with limited data compared to ANN on many occasions. The result, after using three different kernels in SVM such as Linear, Polynomial, and Radial Base Function kernel (RBF), shows that the RBF kernel produces better average accuracy and Cohen’s kappa value compared to Linear and Polynomial kernels, where the average accuracy of each kernel is 78.56% for Linear, 81.40% for Polynomial, and 82.51% for RBF.

Analysis of Graphomotor Tests with Machine Learning Algorithms for an Early and Universal Pre-Diagnosis of Dysgraphia

“…DYS/TD Repartition Model Used Score [11] 27/27 Random Forest 96.43% [12] 56/242 Random Forest 97.9% [13] 48/43 SVM (RBF Kernel) 82.51% [16] 24/971 RNN 90% 1 [13] 42/36 Random Forest 67% [14] 42/36 SVM 66% [14] 42/36 AdaBoost 64% [15] 57/63 Random Forest 77.6% 2 [15] 57/63 SVM 78.8% 2 [15] 57/63 AdaBoost 79.5% 2 [20] 122/458 SVM (RBF Kernel) 83%…”

“…Because we have a labeled database, we worked with supervised machine learning algorithms for our study. Several classes of algorithms have been used in the literature [11][12][13][14][15][16]40], therefore 9 models were tested, with different sets of hyperparameters, to select the best ones for each model.…”

“…The aim of this study is to design a pre-diagnosis tool that would help the examiners and give a first insight into the writing difficulties children can have. Several attempts to create such a tool have already been published [11][12][13][14][15][16][17]. They generally involve graphic tablets that are able to record the path of the pen as a function of time, and usually a classification model that will do a prediction task, with the notable exception of [16].…”

Development of a Pre-Diagnosis Tool Based on Machine Learning Algorithms on the BHK Test to Improve the Diagnosis of Dysgraphia

“…SVM is one of the most commonly used Machine Learning technique [16]. It is known to perform well even with high dimensional data and small datasets [30].…”

“…Most of the studies about dysgraphia are based on small population, around 30 or 40 children [5,6,[14][15][16] and have a descriptive approach, trying to determine whether a variable seems discriminant or not. Our approach is different as we mainly use the features previously demonstrated as discriminant to predict if a child has dysgraphia or not.…”