Effects of noise type and signal-to-noise ratio on speech intelligibility and the speech transmission index.
Stephen Secules · The Journal of the Acoustical Society of America · 2010
The STI metric is the most reliable measure we have for assessing the intelligibility of speech in public address/voice alarm or speech reinforcement systems. A good design for speech is based on an understanding of the room acoustics and electroacoustics of the system, with a consideration for the expected signal to noise ratios in the room. In modeling, most noise signals are assumed to be broadband, whereas in reality the noise signals can be significantly variable in the frequency and time domain, which may dramatically increase or reduce the intelligibility. This study compares the effects of type of noise on intelligibility as measured through both word scores and STI. It explores masking from both theoretical and real world signals. The determined relationships will help inform the design process by clarifying the effects of non-broadband and variable noise signals on intelligibility.