Multilingual speech control for ROS-driven robots
Dominik P. Hofer, Felix Strohmeier (2019): Multilingual speech control for ROS-driven robots In: e & i Elektrotechnik und Informationstechnik, Springer Verlag
To improve the collaboration between humans and robots, multilingual speech control (MLS) can be used to easily manage multiple robots at any time by spoken commands. Once a command is recognised by one of the corresponding ROS-driven robots inside the network, it will be executed and a related audio feedback is provided to the user. Our MLS implementation has a modular design, so that single functional modules can be implemented by either online cloud-based services or by local offline software for increased privacy. Furthermore, the extensible design allows to meet future user needs or to be adapted to different robot capabilities. The MLS follows a principal workflow: Initially, a language identification analysis is done, followed by speech-to-text transformation. Afterwards, the intent is detected and possible variables are analysed for the interpretation of the command, which is furthermore sent to the corresponding robot. Finally, the robot will publish the state achieved by the command execution back to the user. We integrated several cloud-services and open-source implementations based on artificial intelligence technologies and achieved a software framework that is used in a scenario with two different robot systems, a collaborative robot arm and an autonomously moving robot car.
collaborative robotics natural language processing language identification speech-to-text transformation text-to-speech transformation intent detection spoken languages, processing of