Application of Graphological Coincidence Applied in the Field of Speech Therapy in Children with Motor Difficulties

The proposal of an application to measure the graphological coincidence according to the activities proposed in the speech therapy in children with motor difficulties is presented, for which the data from The Quick, Draw! Google Creative Lab dataset is obtained from the game The Quick, Draw! And made available to developers, this base The Quick, Draw! Dataset provides 50 million drawings from 345 different categories. For the project, the database of the circle (contains 118805 drawings of circles), square (120538 drawings of squares), and Triangle (contains 120499 drawings of triangles) was downloaded in the format of simplified drawings in Numpy bitmaps, which contain images 28 x 28 grayscale. The existing database was used to obtain an acceptable level of recognition, with the data obtained, training was carried out with respect to the drawings made by regular children and with motor difficulties using an ergonomic pencil. For the experiment procedure, a certain amount of data is extracted from each figure for training and testing. With 4000 training data and 1000 test data, they are enough to obtain optimal results in the recognition of the 3 figures. Keras was used for the neural network because it has more development tools to use and is optimal for rapid prototyping compared to TensorFlow which is high performance. The Classification Script is used to carry out the process, it uses the Preparation Script to obtain the data from the Training and Test folders by the label and prepares them to be used in the Keras model.