Publication

Studying the performance of automatic speech recognition systems on older adults

Conference Article

Conference

IEEE International Symposium on Robot and Human Interactive Communication (RO-MAN). Workshop in Trends in Socially Assistive Robotics: Human-centered Approach (TSAR)

Edition

2024

Doc link

https://sites.google.com/view/tsar-workshop

File

Download the digital copy of the doc pdf document

Abstract

Automatic Speech Recognition (ASR) systems still struggle in real-world applications, particularly under challenging noise conditions. In this work, we focus on the case of assistive robots interacting with older adult users. We address this gap by creating a novel evaluation dataset that replicates the acoustic challenges encountered in such scenarios. We benchmark the performance of state-of-the-art ASR systems on this dataset. Our results highlight important limitations when the user uses a monotonous tone, or has speech difficulties, or when the robot is far from the user. Thus, user training in the use of the technology is crucial.

Categories

robots.

Scientific reference

C. Escolano, C. Barrue, J. Picas and G. Alenyà. Studying the performance of automatic speech recognition systems on older adults, 2024 IEEE International Symposium on Robot and Human Interactive Communication (RO-MAN). Workshop in Trends in Socially Assistive Robotics: Human-centered Approach, 2024, Pasadena, California, USA.