Evaluación Práctica de las APIs de Procesado de Audio de Microsoft HoloLens 2 para el Desarrollo de Aplicaciones de Realidad Mixta con ASR

Abstract

[Abstract]: Automatic Speech Recognition (ASR) is emerging as a crucial technology for voice recognition in future Extended Reality (XR) and Metaverse applications, enabling natural interactions with virtual and IoT objects. This paper evaluates the performance of Microsoft HoloLens 2 microphone APIs for sound acquisition, quality and noise cancellation in Mixed Reality (MR) scenarios. Three APIs are assessed focusing on sound quality, noise cancellation and speech filtering processes using ASR metrics like Character Error Rate (CER) and Word Error Rate (WER). The provided findings offer insights for MR developers, emphasizing the importance of selecting the right API to enhance sound quality and ASR model precision. This highlights the significance of proper API selection for optimizing MR applications.