Securing social acceptance is essential for the expansion of urban operations of urban air mobility (UAM) vehicles and small unmanned aircraft systems (sUAS). However, conventional exposure metrics based primarily on equivalent sound levels may not fu...
Securing social acceptance is essential for the expansion of urban operations of urban air mobility (UAM) vehicles and small unmanned aircraft systems (sUAS). However, conventional exposure metrics based primarily on equivalent sound levels may not fully explain perceptual differences arising from the temporal and spectral structures of emerging rotorcraft noise. This study investigates the relationships between sound quality metrics (SQMs), a psychoacoustic annoyance (PA) model, and subjective annoyance responses for rotor noise from sUAS and a helicopter, using web-based listening tests. Based on the observation that source identity recognition can influence annoyance judgments even under similar acoustic conditions, a recognition-based correction factor and an extended PA index are further proposed.
First, hovering noise from an sUAS rotor and a helicopter was analyzed by computing SQMs and PA and comparing them with subjective annoyance responses obtained from an online listening test. The results indicate that, under the same sound pressure level condition, the annoyance predicted by the PA model may deviate from listeners’ reported annoyance. In open-ended responses, participants frequently described the helicopter sound with positive or neutral impressions (e.g., “familiar,” “reassuring,” “something to watch”), whereas the sUAS sound was often characterized using negative descriptors (e.g., “insect-like,” “sharp”). These findings suggest that non-acoustic factors associated with perceived source identity and meaning may contribute to the observed discrepancy between PA-based predictions and subjective annoyance.
Second, noise measurements were conducted for the same sUAS platform while varying the rotor blade count (2-, 3-, and 4-blade) under a constant thrust condition. The directivity of SQMs and PA was examined across observation angles. In addition, level scaling was applied to the noise measured at −60° to generate five sound pressure level conditions (60 ~ 80 dB) for the 2- and 4-blade rotors, followed by an online listening test. In certain level ranges, the trend in subjective annoyance with blade-count variation was found to be consistent with a representative PA-based value, indicating that within the same source category, PA can reflect relative changes in “signal-intrinsic” annoyance to a meaningful extent.
Finally, motivated by the potential PA–subjective mismatch observed in the cross-category comparison, this study proposes a minimal correction framework based on observable recognition outcomes. The proportion of responses that explicitly mentioned the correct source label in open-ended descriptions is defined as a proxy for identity recognition (correct label mention rate,p_(correct) ). To address the possibility of zero observations under finite sample sizes, a smoothed mention probability (tilde(p)) is introduced, which is then mapped to a bounded, monotonic familiarity/recognition correction factor F_(fam). An extended index is defined as PA_(ext), where PA_5 represents a intrinsic annoyance component of noise source. The proposed framework does not aim to establish a regulatory equation or a high-accuracy predictive model; rather, it provides a data-reproducible procedure and example calculations for incorporating a minimal recognition-related pathway that may modulate annoyance judgments in cross-category comparisons.