Human-centric Maintenance Process Through Integration of AI, Speech, and AR

TL;DR

This paper presents an integrated AR, AI, and speech recognition system for industrial maintenance, demonstrating how immersive, hands-free interactions can improve efficiency, safety, and reduce cognitive load in maintenance tasks.

Contribution

It develops a demonstrator using Microsoft HoloLens 2 that integrates speech-to-text and AI to support real-time, hands-free maintenance interactions, advancing human-centric AR applications.

Findings

High speech transcription accuracy achieved in AR environment

Enhanced natural interaction between users and augmented systems

Potential reduction in cognitive load and workflow streamlining

Abstract

The adoption of Augmented Reality (AR) is increasing to enhance Human-System Interaction (HSI) by creating immersive experiences that improve efficiency and safety in various industries. In industrial maintenance, traditional practices involve physical documentation and device interactions, which might disrupt the task, affect efficiency, and increase the cognitive load for the maintenance personnel. AR has the potential to support and enhance industrial maintenance processes in these aspects. Therefore, the purpose of this research is to study and explore how advanced technologies like Artificial Intelligence (AI), AR and speech processing can be integrated to support hands-free, real-time task logging and interaction in maintenance environments. This is done by developing a demonstrator for Microsoft HoloLens 2 using Unity, C#, Azure Cognitive Services, and Azure Functions, which enables speech-to-text transcription for hands-free maintenance support. Using Azures' speech recognition, the demonstrator can achieve high transcription accuracy in an AR environment, facilitating natural interactions between users and the augmented environment. The study aims to explore the potential of AR to reduce cognitive load, streamline workflows, and improve safety by enhancing HSI for maintenance personnel in high-stakes environments.