AVASpeech-SMAD: A Strongly Labelled Speech and Music Activity Detection Dataset with Label Co-Occurrence

TL;DR

AVASpeech-SMAD is a new open-source dataset with detailed polyphonic labels for speech and music, aiding research in activity detection with improved annotation quality and benchmark results.

Contribution

It introduces the first open-source dataset with strong polyphonic labels for both music and speech, extending AVASpeech with frame-level annotations.

Findings

Dataset contains 45 hours of audio with manual annotations.

Evaluation of two state-of-the-art SMAD systems provided benchmarks.

Automatic quality checks enhanced label accuracy.

Abstract

We propose a dataset, AVASpeech-SMAD, to assist speech and music activity detection research. With frame-level music labels, the proposed dataset extends the existing AVASpeech dataset, which originally consists of 45 hours of audio and speech activity labels. To the best of our knowledge, the proposed AVASpeech-SMAD is the first open-source dataset that features strong polyphonic labels for both music and speech. The dataset was manually annotated and verified via an iterative cross-checking process. A simple automatic examination was also implemented to further improve the quality of the labels. Evaluation results from two state-of-the-art SMAD systems are also provided as a benchmark for future reference.