Usability of voice-controlled product interfaces

Voice-controlled operation of an audio set was evaluated in a quasi home environment with 76 subjects, who had free choice between voice input and conventional remote control. The set consisted of a triple CD changer, a double audio cassette player, and a tuner. Evaluation was divided into four studies, examining different usability issues. Studies 1 and 2 focussed on the comparison of a perfect speech recognizer (Wizard-of-Oz simulation) with two state-of-the-art automatic speech recognizers (ASR). In studies 3 and 4, different types of feedback were compared: spoken feedback, non-speech audio feedback, and no feedback. The main conclusion is that users are positive about voice-controlled interfaces with feedback. However, it is clear that ASR performance has a major influence on users appreciation of voice controL AlIowing users to adapt some of the interface design options depending on the situation of use contributes greatly to the success of such interfaces. Furthermore, fine tuning many of the design decisions of a voice-controlled interface, such as the type of feedback, feedback sound, and vocabulary design, is especially important as long as state-of-the-art ASR technology still is far from perfect