Fraunhofer IDMT develops audio technology for Deutsche Telekom’s intelligent personal assistant
The Fraunhofer IDMT in Oldenburg has developed the audio technology for Deutsche Telekom’s smart speaker. At the core is the optimized interaction of speakers and microphones for voice control even under difficult ambient conditions. Besides that, the IDMT has achieved outstanding acoustics for telephony and music playback in a small enclosure. The intelligent personal assistant aims to make everyday life easier through voice control and connected services.
Researchers at the Fraunhofer Institute for Digital Media Technology IDMT in Oldenburg are making a special user experience possible through their audio design of the Deutsche Telekom’s smart speaker. By refining and combining microphone and loudspeaker technologies, they have created a very robust voice control system that allows the user to enter speech commands and listen to music at the same time.
“It greatly helped the overall result that we - as audio experts - were involved in the project from the outset. Through controllable directional microphones and signal enhancement techniques for better recognition of voice commands, we were able to give the device superb ‘hearing’. This is also clearly noticeable when using the integrated hands-free function to make phone calls. As far as audio design on the output side is concerned, we have succeeded in developing and aligning all acoustic components in such a way that we can produce an excellent sound within the small dimensions of the enclosure,” says Jan Wellmann, Head of the Audio System Technology Group at the Fraunhofer IDMT in Oldenburg.
“The results produced by the Fraunhofer researchers in Oldenburg make your ears prick up: The smart speaker understands voice commands from a larger distance, in a noisy environment or when music is playing,” says Gunnar Mayer, the Deutsche Telekom project manager responsible for the new speaker. “What makes this possible are the acoustic design of the compact high-end stereo audio system with an output of 25 watts together with a combination of four high-sensitivity microphones optimized for the highest possible level of voice control precision,” continues Mayer.
Many interfaces and high level of data privacy
With the self-learning personal assistant, voice control can be used to operate connected devices in the home. For example, Deutsche Telekom’s TV services and smart home applications can be operated without a remote control. Handsfree Telephone calls are possible too. Other applications are available from cooperation partners such as the news programme “Tagesschau”. Alexa users can also optionally access a large number of Amazon apps with the smart speaker.
As far as data privacy is concerned, the smart speaker complies with the highest German and European data privacy legislation requirements and this is exactly what sets it apart from the competition. All data processing takes place at Deutsche Telekom’s high-performance data centres. Voice data are processed in anonymous form in the cloud and deleted after 30 days.
Project Group Hearing, Speech and Audio Technology at the Fraunhofer Institute for Digital Media Technology IDMT in Oldenburg
The objective of the Project Group Hearing, Speech and Audio Technology at the Fraunhofer IDMT is to transpose scientific findings in the field of hearing perception into technological applications. Its applied research priorities are the improvement of sound and speech intelligibility, personalized audio reproduction and computer-based acoustic speech and event detection. Through scientific partnerships, the Fraunhofer IDMT has close links with the Carl von Ossietzky University, Jade University and other institutions in Oldenburg engaged in hearing research. The Fraunhofer IDMT is a partner in the »Hearing4all« Cluster of Excellence.