Speech recognition as part of automatic decision-making systems has advanced in recent years, becoming a consistent reality in several engineering sectors. Especially in Power Systems, Speech-to-text recognition allows a significant increase in the quality of process operation involving audio communication. In this way, it becomes possible to transcribe the audios involving the communication and also future audits. A growing number of tools have been proposed lately in order to automate speech recognition, but these still have analysis limitations, not allowing redundancy in the transcription process, for example. This work proposes a methodology for analyzing audio in separate channels based on recordings of calls between electrical system operators, aiming increasing the degree of robustness in the application of speech-to-text recognition processes.