Further Validation of the Speech Transmission Index (STI)
The present article further evaluates the accuracy of speech-recognition predictions made according to two forms of the Speech Transmission Index (STI) for normal-hearing listeners. The first portion of this article describes the application of the modified Speech Transmission Index (mSTI) to an extensive set of speech-recognition data. Performance of normal-hearing listeners on a nonsense-syllable recognition task in 216 conditions involving different speech levels, background noise levels, reverberation times and filter passbands was found to be monotonically related to the mSTI. The second portion of this article describes a retrospective and prospective analysis of an extended sound-field version of the STI, referred to here as STI x . This extended STI considers many of the variables relevant to sound-field speech recognition, some of which are not incorporated in the mSTI. These variables include: (a) reverberation time; (b) speech level; (e) noise level; (d) talker-to-listener distance; (e) directivity of the speech source; and (f) directivity of the listener (eg., monaural vs. binaural listening). For both the retrospective and prospective analyses, speech-recognition was found to vary monotonically with STI x .