Improving Speech Recognition on Augmented Reality Glasses with Hybrid Datasets Using Deep Learning: A Simulation-Based Approach

A study presents a novel approach to improving speech recognition on AR glasses in noisy and reverberant environments. The proposed method uses a hybrid dataset of real-world and simulated impulse responses to enhance model performance. Simulated data allows for cost-effective generation of diverse acoustics data.