Spoken Words Prediction Using Convolutional Neural Network for Dysphasia Patients

Abstract

Dysphasia is a medical issue caused by brain damage. This condition affects the ability to produce and understand spoken language. People with dysphasia often have difficulty with verbal communication with others. As a result, this area has an attractive place in technology development to implement a system to help those people to communicate easily with their community. In this paper, one type of dysphasia status is considered which expressive dysphasia is. This type has many symptoms including striving to find words, speaking difficulty, removing small words, using nonsensical words, etc. The aim of the work is implementing a system to understand and predicate the spoken words which are pronounced by the dysphasia patients. The proposed method is based on using Convolutional Neural Network (CNN) to estimate the correct ones with minimum error rate. The main steps of the proposed system contain three vital steps: the first one is gathering data for patients with different ages. Secondly, training the system to learn the classes of the words. Finally, recognize the spoken words with correct estimation.  The prediction rate is 99.41% which is significant and promises results to consider this method to use for patient people to help them to communicate easily with others.