A deep learning-based smart application for malay vowel recognition toward rehabilitation treatment

A stroke is a medical condition in which poor blood flow to the human brain may cause by several factors of cell death. In general, recovery and treatment of stroke patients require an extended time frame consisting of consistent medication and training or rehabilitation training. However, this rehabilitation process is still conducted manually in most rehab centres over the globe. Since the impact of computer vision in this science and technology over the last two decades, machine learning and deep learning have been injected into this medical field to assist rehabilitation via the new technology. This paper introduces the simple proposed neural network with a graphical user interface to help and improve rehabilitation activities. This proposed project aims to create a neural network that can recognize vowels from the voices of a mixed average healthy person with samples from stroke patients. The proposed network method is ideal for training and validating vowel recognition accuracy by utilizing a simple dataset of both groups of people. The proposed project's result presents an intelligent application of the Convolution Neural Network for Malay vowel detection with good performance accuracy. With simple setting, an epoch of 20 and a batch size of 6, the maximum accuracy was 92.96 %. This result reflected that the proposed method is still comparable to other comparative methods with only a simple network design and a limited range of image datasets.