Emergency Stop System of Computer Vision Workstation Based on GMM-HMM and LSTM

Abstract

Voice recognition and command technology for applications with industrial robots is a relatively new field in the intelligent manufacturing industry. It offers a number of advantages over other methods of communication with robots, as it requires fewer specialized skills to manipulate the robot workstation. Additionally, using voice commands can help reduce the number of industrial injuries caused by contact with machinery, thus potentially save operators' lives in emergency situations where external assistance is not immediately available. This study presents a design of a Cartesian robot workstation which is equipped with a voice recognition system controlled by audio commands, as well as a vision perception system. The vision perception system uses the Real Sense depth camera that captures information about the coordinates of the work pieces, which is processed by SSD algorithm. The voice recognition system has been developed with an algorithm which combines both LSTM and HMM, and it has good performance in term of both efficiency and accuracy in controlling normal operation as well as emergency stop for our robot grasping workstation.