Robust Real-time Automatic Voice Command based on Raspberry pi for assistance disabled people

Abstract

Walkers are widely used by people with limited mobility. Much research work is underway to design human-machine interfaces (HMIs) using physiological signals to improve the control of mechanical mobility devices, mainly wheelchairs. Generating exact control commands through physiology in a suitable HMI is a real challenge because severely disabled people cannot control classic wheelchairs. In this context, this work develops a new voice-signal control system to meet the needs of this physically disabled population. This system is divided into two parts: part one covers the recognition, processing and classification of speech signals in this part the relative spectral-perceptual linear prediction (RASTA-PLP) and discrete wavelet transform (DWT) are combined to process and extract speech features and the second part includes the mechanical collection module, i. e. the wheelchair control by a motor drive circuit. The microphone servers for real-time voice recording. A Raspberry-Pi with a Linux operating system kernel is used as the processor. In order to make the processor more user-friendly and reliable, Voice-received mode is integrated into the Wheelchair. The model works successfully with an average recognition of 100 in clean environment and between 80 and 100 in noisy environment.