DescriptionWe present a speaker-aware robotic system which recognizes users by voice in realistic, noisy conditions, highlighting the potential of speaker identification to enrich industrial and social HRI. We approach this as a CNN-based audio classification task, with the particular aim of producing fast, reliable, and explainable predictions. Our method is evaluated on a challenging 6-speaker dataset collected "in the wild" and showcased in a manufacturing scenario, where a collaborative robot personalizes its responses and prevents non-authorized users from executing commands.
|Period||9 Aug 2021|
|Event title||30th IEEE International Conference on Robot and Human Interactive Communication|
|Location||Vancouver, Canada, British Columbia|
|Degree of Recognition||International|
Documents & Links
Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review