Real-time voice-controlled human machine interface system for wheelchairs implementation using Raspberry Pi

Abstract

The article describes the development of a wheelchair prototype designed to facilitate the mobility of people with disabilities. The proposed system is based on voice commands to ensure communication between humans and machines. The system consists of two modules. The first module involves the detection, processing and classification of actual voice signals acquired from a mobile phone. This module incorporates a robust and excellent speech recognition strategy. Indeed, the combination of Mel Frequency Cepstral Coefficients (MFCC) and the Discrete Wavelet Transform (DWT) in signal processing and feature extraction allows for better performance, achieving an effective recognition rate of 100% for an SNR of 5 db. The second module presents the mechanical design and development of the actual prototype, which enables real-time simulation of the first module. This module is based on a Raspberry Pi 3 board with a Linux operating kernel. Finally, tests carried out on the designed wheelchair prototype have demonstrated its efficiency, robustness and excellent response to critical situations, particularly obstacles.