“Bag 3 kg of hex screws and take them to the packaging department”– what used to be a typical instruction to be executed by an employee working in a logistics department will increasingly be a command directed towards a voice controlled robot. But does the robot know which type of screws it is supposed to bag? Yes, it does – because it features integrated acoustic event detection. 

At Hannover Messe, the experts of Fraunhofer IDMT will be demonstrating what voice controlled human-machine interaction can look like. The solution to be presented is particularly interesting for being used in industrial plants and production facilities, as it allows the use of speech recognition also in noisy environments and in settings in which the speaker and the machine to be controlled are located at some distance from each other. The demonstration will feature an industrial robot platform as it is used in automated production processes. It will be shown how the system is able to recognize and distinguish speech and other sounds and how it follows commands accordingly. 

The system features robust speech recognition technology based on state-of-the-art scientific knowledge from psychoacoustics and psychophysics. Furthermore, the system is able to take acoustic conditions of the environment into account, such as background noise or reverberation. While the vocabulary used for speech control basically can be chosen at will, users of the system should be aware that the more the vocabulary is specified to the requirements of the individual application, the better the machine is able to recognize the commands and act accordingly. “The speech recognition we use in our system can be adapted to cover the full range, from very simple voice based machine control using only a few words to sophisticated dialog based control of large robotic systems”, says Dr. Stefan Goetze, co-developer of the system.  

Apart from speech recognition and voice control, the intelligent algorithms developed by Fraunhofer can also be used for acoustic process monitoring (to oversee filling or bottling processes, for example) and acoustic assessment of products or materials. 

In another demonstration, the Fraunhofer researchers will be showing how bulk material can be identified during the filling process based on its acoustic features (which is important to make sure that the right material is used in an automated production process). 

If you are interested in Fraunhofer IDMT’s solution for voice based machine control, acoustic monitoring of processes, and acoustic identification of materials, please feel free to visit us at the Fraunhofer booth C22 in hall 2.