Selective Speaker Interface to Support Smart Space Accessible Computing

The Smart Flow system developed by NIST provides a platform to develop standards that promote the interoperability of devices produced by different manufacturers. The NIST system can acquire and process multiple sensor data such as voice and image in real-time. Many applications such as speaker verification, head trackers, etc. can be implemented under the framework. In Phase 1, we have demonstrated that a speaker can be correctly identified in real-time from speech acquired through a microphone array. In Phase 2, we propose further enhancements to the speaker interface with the aim of achieving robust and accurate speaker verification and speech recognition for speakers with special needs in noisy environment. In particular, we will: (i) acquire speech continuously by establishing plausible voice tracks; (ii) perform adaptive beam forming along voice tracks; (iii) perform speech segmentation based on speech utterances; (iv) further improve speech verification by using utterances rather than a fixed number of speech frames; (v) establish good communication interface between PDA and Smart Flow system; (vi) carry out two real-time demonstrations to illustrate the improved speaker interface; (vii) transfer some of the technology in the Smart Flow system to bird monitoring and classification.