야외 군중 소음 환경 내 신호대잡음비에 따른 비상 방송 인지 성능 평가

Abstract

In dense urban environments, large-scale outdoor gatherings such as festi vals, demonstrations, and public celebrations frequently expose crowds to ext reme background noise levels. These conditions critically hinder the effective ness of emergency communication systems, particularly voice-based announc ements designed to deliver evacuation guidance during crises. While current regulations for emergency broadcasting are predominantly based on indoor e nvironments, they often fail to account for the auditory complexity of real-w orld crowd settings. This study addresses this gap by empirically evaluating the perceptual thresholds of emergency voice broadcasts under realistic outdo or noise conditions and proposing performance-based guidelines grounded in signal-to-noise ratio (SNR) analysis. To replicate authentic crowd noise scenarios, background audio was recor ded at the 2024 Gwangalli Fireworks Festival, a major national event attende d by over one million people. Two representative background noise condition s (75 dB and 97 dB) were extracted from the recordings. These were combin ed with emergency voice broadcasts of varying intensities to produce six dis tinct experimental conditions, corresponding to SNR levels of +5 dB, +10 dB, and +15 dB at each background noise level. A total of 33 healthy adult partic ipants took part in the experiment, where they navigated predefined routes o n a university campus while listening to pre-recorded emergency instruction s through noise-canceling earphones. All participants were exposed to all six SNR conditions using a repeated-measures design. Behavioral metrics included recognition success (i.e., whether participants correctly interpreted the spoken evacuation instruction) and the time taken to initiate evacuation after hearing the broadcast. Data were collected through wearable GoPro cameras and mobile tracking applications. Statistical analysis showed a clear and significant decline in recognition rates at lower SNR lev els, particularly under the 97 dB background noise condition. Even at +5 dB SNR, recognition was frequently incomplete or delayed, suggesting that the i ntelligibility threshold under high-noise outdoor conditions is considerably hi gher than current assumptions. Only at +15 dB SNR did recognition rates co nsistently exceed acceptable thresholds for reliable evacuation guidance. These findings underscore the necessity of adjusting emergency broadcast volume in proportion to real-time ambient noise, rather than applying static output standards. Specifically, the results support establishing a minimum S NR of +15 dB under high-noise (≥ 95 dB) conditions to ensure intelligibility. The study highlights the shortcomings of current regulations, which largely disregard acoustic interference and human perceptual limits in dynamic crow d environments. This work presents a validated experimental framework for evaluating voi ce broadcast effectiveness in realistic crowd noise conditions. By quantifying the perceptual boundaries of emergency voice communication and their behav ioral implications, the study offers foundational data to support the developm ent of revised, performance-based emergency broadcasting criteria for outdoo r public settings in Korea and beyond.