Calculating speech intelligibility for the design of public address systems at railway stations

Abstract

Station public address (PA) forms part of a wider passenger information system that otherwise provides visual information. The PA system makes a unique contribution as information is actively broadcast rather than passively displayed and people's attention, including those with impaired vision, is thereby attracted. This valuable contribution is lost in cases where the announcements are unintelligible. A method is presented for the calculation of the speech transmission index (STI) of speech intelligibility in railway stations; this method may be used to design new PA systems. The method uses a quasi-diffuse model to calculate the level of PA announcements broadcast over a given physical area, the modulation transfer function and finally the STI. The method is shown to produce predictions of the rapid speech transmission index (RASTI) that agree well with results measured on railway station platforms in the open air and on platforms in deep-tube underground stations. The method is described as being hybrid empirical/deterministic as it requires prior knowledge of the spatial average reverberation times and of the prevailing background noise levels while calculating other parameters. This approach along with simplifying assumptions for the relative influence of direct and reverberant speech allows the rapid calculation of STI or RASTI and removes the need to describe the station structure in three dimensions as required by completely deterministic methods. The common requirement for PA systems to meet a performance target of at least 0.5 STI is investigated for station platforms in the open air and those in deep-tube underground stations.