Mapping vocal health risk for active vocal users across European urban soundscapes: a computational framework with exploratory empirical evaluation.

Urban soundscape assessments focus on the listener-the person who hears an environment-while largely overlooking the person who must produce sound within it. However, speaking or singing in noisy settings imposes substantial physiological demands due to Lombard effect, which causes vocal intensity to increase as ambient noise raises. In this study, we present a computational framework that integrates the Lombard effect, subglottal pressure estimation, and signal-to-noise ratio (SNR) to translate urban noise levels into quantitative estimates of vocal load. Calibrated against professional-singer data, the subglottal-pressure estimates reproduce independently measured sound levels to within 0.7 dB. This framework has been applied to strategic road-traffic noise reported under the European Environmental Noise Directive, mapping vocal health risks across urban areas in Denmark, the Netherlands, and France. In these regions, high-risk zones account for 4.3-5.6% of the noise-affected area; an exploratory analysis of 16 singers recorded in two environments is directionally consistent with the framework's central assumption. The framework's principal finding is a divergence between its two risk indicators. As vocal compensation strengthens, SNR-based risk falls, while estimated subglottal pressure rises. This means that successfully compensating for sound typically raises audibility and adds to the physiological load, so an assessment based only on SNR indicates improvement at the moment the hazard increases. Communication-based metrics can therefore conceal, rather than reveal, rising vocal load, and should not be used in isolation. More broadly, the results argue for extending the human-centered soundscape paradigm to those who actively produce sound in the city, rather than only to those who receive it.