A Real-time Voice Interface for Intelligent Wheelchairs

Abstract

This paper reports on the development of a real-time voice interface for navigation purposes of electric wheelchairs. To this end, we employ a convolutional neural network trained and fine-tuned using a small dataset that consists of Greek commands. Furthermore, the study explores a highly quantized version of the network to achieve computational efficiency while maintaining high accuracy on an edge device. The experimental results confirm the effectiveness of the model in accurately detecting keywords in real time with minimal misclassifications.