The Design of Chinese Speech Recognition Interface for Home Entertainment System and Usability in Multiuser Environment

Experiments were conducted to assess a proposed multi-user Chinese speech interface for a home entertainment system. The multi-user speech interface allows designation of speech input to only one or to all users. Tasks were designed to stimulate cooperative and competitive behavior under time constraints. A total of 12 groups were tested. Each group consists of three subjects, who are either friends or members of the same family. The whole experimental process was recorded on video and analyzed with graphic representations. Our results indicate that even for competitive tasks, conflicts among users in the shared mode can be resolved in at most two turns. We also found that recognition performance is a critical influential factor for the usage behavior. In total, 49% of the control switches from one subject to another were caused by recognition errors. Most subjects preferred the master mode to the hands-free shared mode due to better recognition rates, indicating that the hands-free advantage for voice control is beneficial only if the recognition rate reaches a satisfactory level.

A home DVD player is implemented with our proposed interface and for related usability studies.

This research is published in the international conference HCI 2001.

S. H., Hsu, L.L., Chen, Y.C., Chu (2001). Designing multiuser voice control interface for home entertainment systems . In C. Stephanidis (Ed.), Universal Access in HCI, 2001, pp13-17. Lawarence Erbaum, New Orleans , USA .

This project was collaborated by Philips during 1999-2000 and was also submited as my master thesis in the graduate school of design in NTUST; related concepts are patent-pending by Philips Research.
2007-2008 Xocial Research in Laboratory CEDRIC CNAM . Paris