Speech playback geometry for smart homes
Ian Vince McLoughlin, Zhipeng Xie · 2014
Smart homes are environments that intelligently communicate with users, primarily through speech, to control various devices and activities relating to daily living. Both automatic speech recognition and text-to-speech are enabling technologies, however the basic home environment is often acoustically noisy and unconstrained in usage, leading to significant practical difficulties with both key technologies. In addition to interfering noise problems, a home occupant may not always be located physically close to either a microphone or loudspeaker. Speech intelligibility in a smart home depends on many factors including background noise, speech power and content. It is also affected by the orientation of the listeners' ears with respect to the speech source. This paper describes a simple technique to adjust perceived phase angle, and reports results from experiments to assess how the intelligibility of speech heard in such an environment will vary with respect to head position.