Voice Signal-Based Wheelchair Navigation Control Using Convolutional Neural Network for Cerebral Palsy Patients

This research applies the use of a convolutional neural network (CNN) in speech recognition to control a wheelchair, with a focus on cerebral palsy patients. In the context of the mobility and independence of such patients, speech recognition becomes an alternative solution to traditional joysticks that may be difficult to use by patients with hand impairments. In an effort to improve accessibility and quality of life for patients, this research proposes the use of CNN technology to efficiently detect voice commands. Through the evaluation of the CNN model using two different testing stages. The results show significant accuracy levels, with offline testing achieving optimal accuracy of up to 100% on validation data. Although testing with test data revealed a decrease in accuracy to approximately 95.70%, real-time testing showed an average accuracy of 74%.