Design and implementation of an automatic speech recognition based voice control system

EasyChair Preprint 5116

Abstract

This paper deals with the study, design, and implementation of an automatic speech recognition-based voice control system for small vocabulary isolated words, the aim is to control the motion of a small mobile robot. The speech recognition algorithm was implemented using Mel frequency cepstral coefficients (MFCC) for feature extraction and the dynamic time warping (DTW) for pattern matching. The NI MyRIO-1900 embedded module and LabVIEW are used for the real-time implementation; it is built around an interactive Human Machine Interface (HMI). The system is designed to recognize eight commands, including a wakeup word. The developed ASR based control system has the ability to learn new commands, and to adapt to new users and even new languages. This system has an improved ability to recognize and execute commands with high accuracy in real-time. It was able to achieve a recognition accuracy of 99.75% for one speaker using a dictionary of 10 word sets and an accuracy of 96.375% for a set of three different speakers using a hybrid dictionary with 10 word sets, when the word sets in the hybrid dictionary is raised to 30, the designed system is able to achieve a recognition accuracy of 99.5 event with three different speakers.

Keyphrases: Dynamic Time Warping, MFCC, NI myRIO, mobile robot, speech recognition, voice control

@booklet{EasyChair:5116,
  author    = {Adnene Noughreche and Sabri Boulouma and Mohammed Benbaghdad},
  title     = {Design and implementation of an automatic speech recognition based voice control system},
  howpublished = {EasyChair Preprint 5116},
  year      = {EasyChair, 2021}}