The idea of having voice based input and output on wearable devices is not simply achieved, by deploying the speech recognition and synthesis engines to such a device and run it. The limitations of the device, especially computational power and memory, pose strong limitations that cannot be handled by conventional implementations. As a workaround, most of the research utilizes a speech service running in the network, and employing traditional client server architectures. The fact that there may be multiple audio input and output devices, and multiple text-to-speech engines and speech recognizers with different capabilities in a given environment remains untouched in most of the current techniques and is the focus of our research towards a ubiquitous computing speech API “Mundo Speech API”.
- Dr. Dirk Schnelle-Walka
- Stefan Radomski