"OK Aura, Be Fair With Me": Demographics-Agnostic Training for Bias Mitigation in Wake-up Word Detection

TL;DR

This paper investigates demographics-agnostic training methods to reduce bias in wake-up word detection, demonstrating significant improvements in fairness across sex, age, and accent groups using the OK Aura dataset.

Contribution

It introduces label-agnostic training techniques, including data augmentation and knowledge distillation, to mitigate demographic biases in wake-up word detection models.

Findings

Demographic bias reduced by up to 83.65% in age groups.

Data augmentation and knowledge distillation improve fairness.

Significant bias reduction compared to baseline models.

Abstract

Voice-based interfaces are widely used; however, achieving fair Wake-up Word detection across diverse speaker populations remains a critical challenge due to persistent demographic biases. This study evaluates the effectiveness of demographics-agnostic training techniques in mitigating performance disparities among speakers of varying sex, age, and accent. We utilize the OK Aura database for our experiments, employing a training methodology that excludes demographic labels, which are reserved for evaluation purposes. We explore (i) data augmentation techniques to enhance model generalization and (ii) knowledge distillation of pre-trained foundational speech models. The experimental results indicate that these demographics-agnostic training techniques markedly reduce demographic bias, leading to a more equitable performance profile across different speaker groups. Specifically, one of the evaluated techniques achieves a Predictive Disparity reduction of 39.94\% for sex, 83.65\% for age, and 40.48\% for accent when compared to the baseline. This study highlights the effectiveness of label-agnostic methodologies in fostering fairness in Wake-up Word detection.