Improving Intelligibility of Airport Terminal Public Address Systems (2017)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

18 3.1 Introduction Although the acoustical requirements for airports are much less demanding than they are for other large public spaces (such as concert halls), the same design tools are available to deliver acoustical success in airports as well. A concert hall is an acoustical success when attention to the details of the acoustical environment results in a high quality of sound. An airport is an acoustical success when passengers can hear and understand announcements that are relevant to them. The attention paid to acoustics in airports has often not been as focused as it could be on important parameters for speech intelligibility, and airport planners have often not taken full advantage of the tools available to them. Some airports have favorable acoustics and an intel- ligible PA system, but in other airports, intelligibility could be improved. Audibility (the fact that sound can be heard) does not necessarily result in intelligibility. Inter- national Electrotechnical Commission (IEC) Standard 60268 defines intelligibility as “a measure of the proportion of the content of a speech message that can correctly be understood.” The speech signal can be degraded in some ways, thus limiting the transfer of information content. A loud PA signal can be unintelligible if the space is too reverberant. Reverberation is the per- sistence of sound in a space, and a highly reverberant space is one with hard surfaces and little acoustical absorption. The result is sound that continually reflects around the space, rather than being absorbed quickly at the room surfaces; sound lingers (slowly decays) and masks succes- sive sounds. To achieve good speech intelligibility, the sound of each word must decay rapidly; otherwise, successive words will be muddied by the lingering sound. A high level of background noise can also interfere with intelligibility by masking the spoken sound unless the spoken sound is sufficiently amplified above the background noise. A good example of this is the so-called “cafe effect” in which many people in a group are talking at the same time, forcing all speakers to raise their voices to be understood, in turn making it even harder for everyone to be understood. A related phenomenon is called the “Lombard effect,” in which speakers modify their normal speech pattern to adjust for an increase in ambient noise. In this case, speakers automatically and subconsciously raise their voices, increase pitch, and improve articulation, resulting in improved intelligibility. 3.2 Background The scientific aspects of speech intelligibility have been studied for many decades. Early tests would either have listeners assign a numerical value or a subjective term to rate how well they could understand a speaker, or they would have listeners document what words or sentences they heard (or thought they heard), so that by experiment an objective number and percentage C h a p t e r 3 Speech Intelligibility

Speech Intelligibility 19 accuracy could be determined. One of the early methods of the second test was a subjective test called the “monosyllabic word intelligibility” test, which used spoken words and a group of listeners to test for intelligibility by measuring the number of words correctly identified. A truly objective method for measuring intelligibility without listeners became available 43 years ago with the introduction of the Speech Transmission Index (STI). The physical parameters that affect speech intelligibility include the method of speech intel- ligibility evaluation; hearing acuity and perception issues; architectural acoustical conditions such as reverberation time, diffusion, and obstructions; background noise; the design of the PA system itself; and announcement quality. The literature review uncovered few documents on this topic specifically relating to airports. Relevant findings are summarized below. Speech intelligibility and special populations • Older passengers benefit from airport spaces with higher STI than those designed for the aver- age passenger (Kim and Soeta 2013; Morimoto, Sato, and Kobayashi 2004; Sato, Morimoto, and Wada 2012). The specific reasons were not controlled, and there may be factors other than hearing impairment for this population. • Hearing-impaired passengers benefit from airport spaces with higher STI than those designed for non-hearing-impaired passengers (Festen and Plomp 1990). • When one language is the only or primary language for announcements, non-native language listeners also benefit from airport spaces that provide a higher STI than those designed for native language listeners (Tachibana 2013, S. J. van Wijngaarden 2001, and van Wijngaarden et al. 2004). Reverberation time • Long reverberation time greater than 1 second is problematic for speech intelligibility (Kim and Soeta 2013, Tachibana 2013). • Speech intelligibility for hearing-impaired and non-native language listeners is most sensitive to long reverberation time (van Wijngaarden et al. 2004, Yokoyama and Tachibana 2013). Background noise • A good target for announcement levels is a signal-to-noise ratio (SNR) of 10 to 15 dB (Morimoto, Sato, and Kobayashi 2004). • At higher SNR in noisy environments, too much signal can degrade speech intelligibility (Morimoto, Sato, and Kobayashi 2004). • Background noise conditions have a disproportionate effect on people with hearing impair- ment and non-native language listeners, and such listeners benefit from a higher SNR (Festen and Plomp 1990, Tachibana 2013). The reader may also be interested in ACRP Report 157: Improving The Airport Customer Experience (Boudreau 2016) and ACRP Project 07-13, “Enhancing Wayfinding for Aging Travelers and Persons with Disabilities.” 3.3 Qualitative Measures of Intelligibility The scientific aspect of speech intelligibility has been studied for over 70 years. Qualitative or—more accurately—subjective measures of speech intelligibility were the initial assessments developed; one such measure is a scale rating the quality of intelligibility. The disadvantage of qualitative tests is primarily one of resources—although a recording can be used to generate

20 Improving Intelligibility of airport terminal public address Systems the test words and sentences, many listeners are recruited to give their assessment. Another dis- advantage is that each listener may have language skills or physical conditions that color his or her ability to offer an unbiased result. Such qualitative test results can be difficult to repeat. One advantage of qualitative measures is that they are useful for understanding issues such as native language or non-native language comprehension and particular sound combinations related to specific languages. 3.4 Signal-to-Noise Ratio The signal-to-noise ratio (SNR) is a measure of how clearly a signal can be heard above noise, and it is a critical factor for speech intelligibility. SNR is defined as the ratio of the information (or signal) over the interference (noise). Given that sound and noise (unwanted sound) are com- monly measured as sound pressure levels (SPLs) using decibels (dB), the ratio of the sound pressures can be equally expressed as the difference in decibels. Industry practice thus uses SNR to quantify the difference between the PA system sound level and the background noise level (e.g., heating, ventilation, and air conditioning noise). On a more basic level, SNR can be viewed as the effect of any unwanted sound that degrades intelligibility, such as sound lingering from announcements due to excessive reverberation. Early research into an objective measure of speech intelligibility focused on the correlation between good SNR values and speech comprehension. Industry practice supports the use of 10 to 15 dB SNR, but a minimum design goal of 10 dB SNR may be adequate, if other positive factors are in place. Figure 3-1 shows the measured SNR at each of the 46 airport spaces measured during unoccupied conditions. The average SNR Figure 3-1. SNR field measured in 46 airport spaces.

Speech Intelligibility 21 was 13 dBA. The main point to understand here is that the preponderance of spaces had an SNR greater than 10 dB. 3.5 Quantitative Measures of Intelligibility 3.5.1 Speech Transmission Index The most widely accepted quantitative measure of intelligibility is the Speech Transmission Index (STI), which is defined in IEC 60268-16:2011, Objective Rating of Speech Intelligibility by Speech Transmission Index. STI values range from 0 to 1, with numbers close to 1 achieving high levels of intelligibility, yet even an STI value of 1.00 is no guarantee that the speech quality heard will be perceived as perfect. This quantitative measurement method relies on comparing a known signal broadcast through the loudspeaker with the sound measured at the receiver (e.g., height of the human ear); the test signal covers the frequency range of human speech with a specific sequence of periodic (repeating) signals. The early research on STI also produced qualitative ratings ranging from “bad” to “excellent” for various ranges of STI values; given that these qualitative ratings are no longer included in IEC 60268, they are not included here. Codes and application standards typically recommend a minimum STI of 0.45 or 0.50. IEC 60268, Annex G, for example, suggests that an STI 0.50 rating is an appropriate “target value for VA (voice address) systems.” The distinction between a VA system and a PA system is that the VA system might be used for emergency or internal use—not for general purpose public messages and announcements. Table 3-1 excerpts information from Annex G of IEC 60268 and includes language about STI value acceptability contained in Annex G. As discussed in Sec tion 3.8, to reach the design target during daytime operations, it will often be necessary to address the fact that daytime ambient conditions are higher than nighttime operations. 3.5.2 Speech Transmission Index for PA Systems The development of instruments to measure STI more efficiently led to the development of the Rapid Speech Transmission Index (“the RASTI method”) in 1979. When RASTI was applied to PA systems, however, shortcomings in the method became apparent. To make it practical to measure of the intelligibility of PA systems, Jan Verhave and Herman Steeneken, using extensive STI Range Typical Uses Comments from IEC 60268 0.66–0.75 Theaters, courts, assistive listening systems, classrooms, concert halls High speech intelligibility 0.62–0.65 Good speech intelligibility 0.58–0.61 Concert halls, modern churches High-quality PA systems 0.46–0.53 Public spaces, cathedrals Acceptable for voice address (target 0.50) 0.42–0.45 Difficult (challenging) spaces 0.00–0.41 Not suitable for PA systems Source: IEC 60268 Table 3-1. Examples of STI qualification bands and typical applications.

22 Improving Intelligibility of airport terminal public address Systems research, developed the STI-for-PA method. United States and international standards exist to define specific measurements for speech intelligibility. Standards also exist for rating the effect of noise on intelligibility. In 1992, a convenient and efficient means of measuring speech intel- ligibility for PA systems, the Speech Transmission Index for PA (STIPA), was introduced. STIPA has come into wide use in the last decade and is an easy means of measuring the STI performance of a PA system in an existing space with background noise (see Figure 3-2). Figure 3-3 shows the STI values from Figure 3-1 in relation to the SNR condition. The STI values measured at all 46 facilities during nighttime or quiet off-hour operations are plotted known signal Test signal Known signalthrough PA Room Room response STI calculaon Figure 3-2. Sound transmission index calculation process. Figure 3-3. STI plotted against SNR, field measured in 46 airport spaces.

Speech Intelligibility 23 against the SNR. The average STI value over all of these measurement locations was 0.51, which addresses the feasibility and practicality of achieving STI performance in conformance with the IEC 60268 VA target of 0.50. There is a general trend of improved STI performance with greater signal-over-ambient conditions. Although the average over all 46 measurement spaces was 13 dB SNR, the achievement of an SNR greater than 10 dB does not automatically result in high STI performance (see Figure 3-3) and other factors are important as well (see Chapter 4). For areas with STI 0.50 or better, the average was 15 dBA, but these STI results were not absolutely fixed to the SNR value. Almost 25% of the spaces measured STI 0.60 or better, and the average SNR was 16 dBA. The highest SNR was not a predictor of high STI performance. Thus, the guidance target is 10–15 dB SNR, in line with the spaces that provided STI 0.50 or better. According to IEC 60268, a 3 dB change in the SNR should result in a 0.10 change in the STI. This is true for a single environment where the only change is the level, but across the 46 field measurements, each with different reverberant conditions and with different frequency charac- teristics, a 3 dB change in the SNR resulted in a much smaller STI change—about 0.03 points. 3.6 Code Requirements The National Fire Protection Association (NFPA) has developed NFPA 72, Annex D (NFPA 2016), which addresses speech intelligibility in specific detail. Although the contents of Annex D are not mandatory, many public agencies use it as a basis of testing for adequate intelligibility. Annex D includes a specific test protocol for voice communication systems, a list of references, terminology definitions, discussion of STI and STIPA, issues having to do with background noise, and acceptability criteria. Annex D describes a clear approach for measuring intelligibil- ity in an existing building and is directly applicable to airport facilities. NFPA 72 also specifies recommended acceptance criteria of 0.50 STI; values as low as 0.45 are acceptable as long as the average performance is 0.50. Annex D also mentions the importance of the design process for new buildings, including that hand calculations are sometimes adequate, but that more com- plex designs are “frequently better and more cost-effectively analyzed using readily available computer-based design programs.” (See Section 7.12 for guidance on practical considerations for combining life safety systems with the general announcement PA system.) 3.7 Other Considerations 3.7.1 Non-native Language Listeners Annex H of IEC 60268-16:2011 indicates that the SNR should be increased by 4 to 5 dBA to provide the same quality of speech intelligibility for non-native listeners; a 3 dBA SNR increase corresponds to a 0.10 improvement in the STI. Given that the guidelines focus on United States airports, the main focus of the discussion of non-native listeners will be on international travelers and on gate announcements and other announcements made in English; however, airports in some regions will want to take this into account for their populations as well. Annex H, Table H.1, indicates that, for non-native listeners, the target STI should be increased by 0.05 to 0.36 above the target goal for native listeners. Table 3-2 is adapted from Annex H, Table H.1, and illustrates the different target adjustments based on language fluency. Table 3-2 also includes some qualitative ratings in association with each STI target value. Practically speaking, it can be challenging to achieve an STI greater than 0.70 in an airport environment, so a cap on the total adjustment to the target goal should be considered.

24 Improving Intelligibility of airport terminal public address Systems 3.7.2 Hearing Impairment, Age-Related Hearing Loss, and ADA Considerations Annex I of IEC 60268-16:2011 provides information on methods to adjust STI targets based on age and general assumptions about hearing impairment. The SNR should be increased by 4.5 dBA to provide the same quality of speech intelligibility for someone with a 20 dB hearing loss defined against the pure-tone average (PTA) hearing level. A 3 dBA SNR increase corresponds to a 0.10 improvement in the STI. The STI is not reliable for all types of hearing impairment, and other researchers use subject-based listening tests or other speech intelligibility methods to predict performance. Age-related hearing loss, however, can be directly adjusted. Table 3-3 is adapted from Annex I and presents these adjusted STI values, showing that the STI should be raised by 0.12 to 0.21 points to account for hearing impairment. As mentioned in Section 3.7.1, a cap on the total STI adjustments should be considered. Passengers with hearing impairments and using conventional hearing aids can benefit from a PA system with a higher SNR setting. However, many airport environments can be noisy, and STI Label Category Standard STI Nonnative Listeners Category I Experienced, daily nonnative language use Nonnative Listeners Category II Intermediate experience and level of nonnative language use Nonnative Listeners Category III New learner and/or infrequent nonnative language use Bad–poor 0.30 0.33 0.38 0.44 Poor–fair 0.45 0.50 0.60 0.74 Fair–good (3) 0.50 0.55 0.68 0.86 Fair–good 0.60 0.68 0.86 Not achievable Good–excellent 0.75 0.86 Not achievable Not achievable Adapted from IEC 60268 Note 1. For details on STI label categories, refer to ISO 9921. Example: To achieve an intelligibility equivalent to an STI of 0.45 for a nonnative Category II listener, the transmission system needs to achieve a performance of 0.60. Note 2. For intermediate values between the stated standard STI, interpolation should be used to estimate the adjusted STI. Note 3. The nonnative category adjustments have been interpolated from the values in Annex H. Table 3-2. Adjusted intelligibility qualification tables relative to standard STI values for nonnative listeners. STI Label Category Standard STI Older listener PTA=15 dB Older listener PTA=20 dB Older listener PTA=30 dB Bad–poor 0.30 0.42 0.47 0.51 Poor–fair 0.45 0.57 0.62 0.66 Fair–good (3) 0.50 0.62 0.67 0.71 Fair–good 0.60 0.72 Not achievable Not achievable Good–excellent 0.75 Not achievable Not achievable Not achievable Note 1. For details on STI label categories, refer to ISO 9921. Note 2. Standard STI values assume that listeners have a PTA between 0 and 5 dB. Note 3. These values have been derived from the values in Annex I. PTA = Pure-tone average hearing level. Table 3-3. Adjusted intelligibility qualification tables relative to standard STI values for listeners over age 60 with hearing loss.

Speech Intelligibility 25 given that noise is also amplified by hearing aids, it is not surprising that people with hearing aids opt to turn them down, relying more often on visual displays instead. Per the ADA (ADA Standards 2010), in each assembly area where audible communication is integral to the use of the space, an assistive listening system shall be provided. In Chapter 7, additional information is provided for integrating audio-frequency induction loops into the PA system. These induction loops are essentially a loop of cable or an array of loops placed around a room or a building to generate a magnetic field that can be picked up by compatible devices such as modern hearing aids. 3.8 Effect of Ambient Noise on STI Given that speech intelligibility and the STI are influenced by the SNR, the ambient noise conditions that affect the SNR also affect the STI. Figure 3-4 summarizes the range of ambient conditions measured, with the average daytime condition being 62 dBA and the average night- time condition being 55 dBA over all 45 spaces where the ambient condition was measured. Note: STI was measured at 46 spaces, but the ambient conditions were measured at only 45 spaces. No ambient conditions were measured at ADS 33. Figure 3-4. Ambient noise levels measured in 45 of the 46 different airport locations.

26 Improving Intelligibility of airport terminal public address Systems Measured daytime ambient conditions were on average 7 dBA higher than the nighttime condi- tions. Section 4.5 presents more discussion on ambient noise, and Section 6.6 presents guidance on controlling these sources. It can be difficult to conduct the STIPA test during daytime operations for some reasons, including interference between the test and operational PA announcements, annoyance of trav- eling passengers, and the potential variability of ambient conditions during daytime operations. Thus, it is important to understand how the STI that is experienced during daytime operations may differ from the STI results obtained during nighttime conditions. The nighttime STI is mea- sured during “dry” conditions, and the in situ daytime STI is measured or calculated during the “wet” condition. Since the SNR during the dry condition is often much greater than 10 dB, the expected STI under wet conditions can often be calculated by adding the daytime or wet ambient to the dry STI measurement result. The expected change in the STI is about -0.20 points. Thus, a dry STI measurement of 0.70 can be expected to result in a daytime effective performance of approximately 0.50 under average conditions. Factors other than the A-weighted ambient sound levels may affect this difference; in particular, since the STI is frequency-dependent, a smaller change (perhaps -0.15 to -0.10) may occur between the dry and wet values, if the daytime ambient conditions can be controlled to maximize the SNR. Ambient noise-sensing systems are becoming more widespread, and they typically provide about 5 to 8 dB additional SNR. An SNR increase of 3 dB can provide a 0.10-point improvement in the STI, but it is also important to know that ambient noise-sensing systems cannot overcome extremely challenging conditions under reverberant or high ambient noise conditions. More details are provided in Chapter 7. 3.9 Guidance Targets 3.9.1 Design • SNR: 15 dBA or better (preferred) in daytime ambient conditions; 10 dBA minimum. This is influenced during architectural design to control the ambient conditions and during PA system design and installation. ADS Use Measured STI (dry) Calculated STI (wet) Difference Nighttime ambient (dBA) Nighttime SNR (dBA) Daytime ambient (dBA) 3 TSA 0.61 0.43 -0.18 51 11 65 4 Concessions 0.65 0.51 -0.14 52 22 62 5 Gates 0.68 0.43 -0.25 52 11 57 6.5 Concessions 0.41 0.15 -0.26 57 11 71 16 Ticketing 0.61 0.53 -0.08 51 21 63 18 Baggage 0.73 0.46 -0.27 47 20 61 19 Baggage 0.63 0.45 -0.18 51 15 63 20 Gates 0.65 0.49 -0.16 58 11 65 21 Concourse 0.49 0.35 -0.14 59 12 64 22 Ticketing 0.52 0.23 -0.29 55 8 66 35 TSA 0.45 0.18 -0.27 56 11 68 43 Curbside 0.39 0.14 -0.25 64 2 71 45 Baggage 0.61 0.50 -0.11 59 15 71 Average -0.20 Table 3-4. Calculated daytime equivalent STI (wet) based on nighttime STI (dry) and daytime ambient conditions.

Speech Intelligibility 27 • Design STI: – Daytime (wet): 0.50. This is the minimum target per NFPA 72, Annex D. – Nighttime (dry): Performance testing and commissioning are done outside normal operating hours. Design for a target STI = 0.60 to 0.70. Specific target value will depend on site-specific conditions. Based on the typical difference between daytime and nighttime ambient conditions, the following can be considered: b 0.60 for a PA system replacement. Without improving the acoustical environment; it may not be possible to achieve much more. b 0.65 based on the project’s ability to control or lower the reverberation time and ambient noise and support the PA system. b 0.70 for new terminal or renovation; many options are available to provide a satisfac- tory acoustical environment to support the PA system. – The design STI can be applied terminal- or project-wide or different spaces can be assigned different STI values. – See Equation 3-1 to determine nighttime (dry) design goal. • Human factors (HF): Add 0.03 to 0.10 points to compensate for challenges that specific passenger populations such as the following may have: – International travelers and non-native language listeners – Passengers with disabilities such as hearing impairment Also be aware of other site-specific considerations for which compensation is required. Equation 3-1. Guidance STI formula during design. = +DryDesign STI 0.60 to0.70 HF[ ] 3.9.2 Performance and Commissioning The STI performance is typically measured during ambient conditions that are lower than conditions during normal daytime operations, and the correction for daytime ambient con- ditions can be approximately -0.20 STI point. If desired, one can develop the performance STI requirement by reducing the design STI by 0.10 to account for an ambient noise-sensing system. • Performance STI: This is the value shown in the specifications. It can be applied terminal- or project-wide or different spaces can be assigned different STI values. See Equation 3-2 to determine nighttime (dry) performance goal. • Ambient noise microphone adjustment (AN): This is an adjustment to allow for ambient noise-sensing in the PA system. The overall performance STI can be reduced by 0.10 point. Equation 3-2. Guidance STI formula to implement in specifications. = -Dry Performance STI Design STI AN