Experimental evaluation of voice conferencing techniques

Abstract

The requirement to support voice conferences in which some participants have only narrow-band speech communication capability prevents the use of the conventional signal summation (bridging) technique. In order to evaluate the relative acceptability of alternative signal selection techniques, a conferencing simulation facility has been constructed, and a series of experiments has been designed and carried out. The effects of conference size, communication delays, and voice encoding techniques have been investigated. Conference control based on speech activity detection has been compared with control using push-to-talk switches or tone keys. Both centralized and distributed control algorithms have been simulated. Test scenarios have been developed that provide both task performance measurements and subjective judgments of conferencing system adequacy. Results indicate that simple broadcast protocols are generally preferred over more complex schemes and that signal selection techniques can perform comparably to signal summation. [This work was sponsored by the Defense Communications Agency for the Department of Defense.]