In this paper, we describe a general methodology for enhancing measurement accuracy in cyber-physical systems that involve structured human interactions with a noisy physical environment. We define structured human interactions as those that follow a domain-specific workflow. The idea of the paper is simple: We exploit knowledge of the workflow to correct unreliable sensor data. The intellectual contribution lies in an algorithm for joint estimation of the current state of the workflow together with correction of noisy sensor measurements, given only the noisy measurements and an overall workflow description. We demonstrate through simulations and a physical implementation the degree to which knowledge of workflow can increase sensing accuracy. As a specific instantiation of this idea, we present a novel situation-awareness tool called the Emergency Transcriber designed to automatically document operational procedures followed by teams of first responders in emergency-response scenarios. Evaluation shows that our system provides a significant fidelity enhancement over the state of the art, effectively coping with the noisy environment of emergency teams.