In a scenario where you need to implement speech-to-text functionality for a call center application that handles customer support calls, what steps would you take to ensure the accuracy of the transcription, especially in noisy environments? Consider the use of Azure services and any necessary customization.