Evaluation of speech intelligibility of sound fields in passenger train compartments

Abstract

Institute for Human Science and Biomedical Engineering,National Institute of Advanced Industrial Science and Technology (AIST),1–8–31 Midorigaoka, Ikeda, 563–8577 Japan(Received 18 February 2009, Accepted for publication 8 April 2009)Keywords: Speech intelligibility, Public address system, Train compartment,Impulse response measurement, Sound fieldPACS number: 43.28.Lv [doi:10.1250/ast.30.379]1. IntroductionThe noise level inside a running train ranges from 60 to80dB for an outdoor ground-level track [1] and 60 to 100dBfor an underground track [2–4]. In such a noisy environment,it may be difficult to understand important announcements forpassengers from public address (PA) systems, such asinformation about travel (e.g., the next station, stations wherepassengers transfer), notices (e.g., the locations of toilets,public phones and designated smoking areas), safety an-nouncements (e.g., warnings of upcoming curves in the track,the locations of open doors), and warnings (e.g., safety withregard to pickpocketing). Therefore, the speech intelligibilityof announcements made through train PA systems isnecessary to relay important information to passengers.The nature of noise exposure experienced by passengersin trains has been previously investigated [1,5]. Environ-mental noises, including the noise of passenger conversations,have been also evaluated from the viewpoint of passengercomfort [6,7]. Although announcements from PA systems andnoises in trains are influenced by the effects that occur in aroom such as the reflection and reverberation of sounds, nostudies have been carried out to specifically investigate thesound fields in trains.Sound sources such as announcements and noises in trainschange as a function of time, while the sound field in eachcompartment does not change. This is because sound sourceshave different acoustical characteristics depending on thespeaker’s voice (or language) and train type, while the soundfield in a compartment is dependent on the compartment’ssize and shape, which are similar amongst all train compart-ments. To investigate the speech intelligibility of announce-ments from PA systems, we analyzed the sound field in a traincompartment. Impulse responses were measured in an emptytrain compartment parked in a depot. The acoustical param-eters extracted from the impulse response were calculated toevaluate the speech intelligibility of the PA system.2. Impulse response measurements2.1. Arrangement of source and receiver positionsA compartment of a train typically used on the suburbanlines in Japan was chosen for the study. The inside dimensionsof the train were 18m in length, 2.6m in width and 2.2m inheight, with seats located against the walls (Fig. 1). Whilevaring the number of sound sources, three types of impulseresponses were measured in the train compartment. In the firstconfiguration (configuration A), three sources and sevenreceiver positions were distributed throughout the compart-ment (Fig. 1(a)). The PA system of the vehicle had sixloudspeakers (PA1 to PA6) located on the ceiling. To ensurethat a loudspeaker located at one end of the compartment (e.g.PA1) did not affect the sound field around a receiver locatedat the other end (e.g. r6), the measurement area was limited tohalf of the compartment. The receiver positions at the front ofthe seats and in the wheelchair user’s area (r1, r4, r5 and r6)were assumed to be representative of the locations of sittingpassengers, while the other receiver positions (r2, r3 and r7)were assumed to be representative of the positions of standingpassengers. Three sources were located under the loud-speakers of the PA system (PA4 to PA6). In the secondconfiguration (B), two sources and four receiver positionswere distributed in the measurement area (Fig. 1(b)). In thethird configuration (C), one source and four receiver positionswere distributed (Fig. 1(c)).2.2. Measurement setupTo represent different PA loudspeakers introduced invarious train compartments, an omnidirectional loudspeaker(Type 4292, B&K) was used. The height of each source was2.0m above the floor. Each receiver was a dummy headmicrophone (KU100, Neumann). For each receiver positionsimulating the location of a sitting passenger, the heights ofthe ear entrances of the dummy head microphones were 1.1mabove the floor. For each receiver simulating the location of astanding passenger, these heights were 1.6m above the floor.The dummy head microphone faced the inside of the vehicle.A sinusoidal signal with an exponentially varying fre-quency swept from 40Hz to 20kHz over a period of 18s wasused for measuring the impulse responses [8]. The two-channel responses at the positions of the left and right ears ofthe dummy head were deconvoluted so that the binauralimpulse responses, i.e., p