Explore chapters and articles related to this topic
Speech and its perception
Published in Stanley A. Gelfand, Hearing, 2017
Another approach to estimating speech intelligibility from acoustical measurements is the speech transmission index (STI), which is based upon the modulation transfer function (MTF). This technique was originated by Steeneken and Houtgast (1980). In addition to their work, the interested reader should also refer to informative sources (e.g., Humes et al., 1986; Anderson and Kalb, 1987; Schmidt-Nielsen, 1987; Steeneken, 2017; van Wijngaarden and Drullman, 2008). An important advantage of the STI is that it accounts for the effects of all kinds of noises and distortions that affect the speech signal, including reverberation and other aberrations that occur over time.
An exploratory investigation of speech recognition thresholds in noise with auralisations of two reverberant rooms
Published in International Journal of Audiology, 2021
Antti Kuusinen, Eero Saariniemi, Ville Sivonen, Aarno Dietz, Antti A. Aarnisalo, Tapio Lokki
The most used measure to quantify the effects of (full) reverberation and noise on speech intelligibility is the Speech Transmission Index (STI; Houtgast and Steeneken 1985). Both reverberation and noise reduce the fluctuations of the speech envelope, and this reduction can be quantified by the modulation transfer function, which is calculated over 14 modulation frequencies (from 0.63 to 12.5 Hz) and seven octave bands from 125 to 8000 Hz. STI can be measured either directly in the room or calculated indirectly from the room impulse responses by considering the sound pressure levels of the signal and the interfering noise as well as the effects of auditory thresholds and masking. Although STI in its basic form does not consider spatial hearing, i.e. the effects of reverberation on the interaural coherence of the masking noise, it has been found to accurately predict speech intelligibility regardless of whether it is disrupted by noise, reverberation or a combination of the two (Duquesnoy and Plomp 1980; George et al. 2010).