The VOiCES from a Distance Challenge 2019
Abstract
The VOiCES from a Distance Challenge 2019 was designed to foster research in the area of speaker recognition and automatic speech recognition (ASR) with a special focus on single-channel distant/far-field audio under various noisy conditions. The challenge was based on the recently released VOiCES corpus, with 60 international teams involved, of which 24 teams participated in the evaluation. In this paper, we separately present the challenge’s speaker recognition and ASR tasks. For each task, we outline the training, development, and test data, as well as the evaluation metrics. Then, we report and discuss the results in light of the participant-provided system descriptions, to highlight the major factors contributing to high performance in distant speech processing.