Academics using intelligibility tests are often interested in talker or listener performance, whereas communications system designers and speech processing algorithm developers are primarily interested in system performance. The latter may find that intelligibility based tests, though more universally applicable than objective tests, yield results that are not readily exchangeable or comparable with those from other workers in the field. This is partly due to the inescapable variations of talker and listener performance, partly to the lack of common standards, and partly to the wide variety of tests available. This paper suggests a test which largely eliminates the data exchangeability problem.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.