Augmenting Highway Workers’ Hearing of Backup Alarm Sounds in a Noisy Workplace Using Speech Enhancement Generative Adversarial Network (SEGAN)

Backup alarm signals are crucial to alert nearby workers of the potential dangers of reversing mobile equipment at construction sites. However, due to the loud ambient noise on construction sites, workers may wear hearing protection devices (HPD), potentially impacting their ability to detect these alarms. There is a lack of an existing framework that can protect workers from loud and complex noise but preserve the audibility of reverse signal alarms. This study addresses the gap by developing a novel framework using Speech Enhancement Generative Adversarial Network (SEGAN) to augment the audibility of backup alarm signals for construction workers exposed to loud noises. The SEGAN model was trained on a large-scale equipment sound data set incorporating 4 types of backup alarms and 26 background noise conditions. We performed evaluations on an independent, unseen test set with varying signal-to-noise ratios (SNR) using objective metrics such as PESQ, CSIG, CBAK, COVL, and SSNR. This model can be utilized to develop a new noise-canceling device, which will have great potential to improve the safety of work zones.