Implementation of a Reliable VUI System on Edge Device

Voice user interface (VUI) is becoming highlighted as an effective tool for customized user interactions. In this paper, we implement a reliable VUI system using speech-to-text technology. The user's utterance is converted into text using Google STT API and followed by keyword determination. Then, the system provides appropriate response corresponding to this keyword query to the user. We use four response types according to voice characteristics (male vs. female) and amount of information (reference vs. short). Our proposed VUI shows a real-time performance on the edge device and has been evaluated on two separate experiments.