Deep multiple instance learning for foreground speech localization in ambient audio from wearable devices

Rajat Hebbar, Pavlos Papadopoulos, Ramon Reyes, Alexander F. Danvers, Angelina J. Polsinelli, Suzanne A. Moseley, David A. Sbarra, Matthias R. Mehl, Shrikanth Narayanan

Research output: Contribution to journalArticlepeer-review

11 Scopus citations

Abstract

Over the recent years, machine learning techniques have been employed to produce state-of-the-art results in several audio related tasks. The success of these approaches has been largely due to access to large amounts of open-source datasets and enhancement of computational resources. However, a shortcoming of these methods is that they often fail to generalize well to tasks from real life scenarios, due to domain mismatch. One such task is foreground speech detection from wearable audio devices. Several interfering factors such as dynamically varying environmental conditions, including background speakers, TV, or radio audio, render foreground speech detection to be a challenging task. Moreover, obtaining precise moment-to-moment annotations of audio streams for analysis and model training is also time-consuming and costly. In this work, we use multiple instance learning (MIL) to facilitate development of such models using annotations available at a lower time-resolution (coarsely labeled). We show how MIL can be applied to localize foreground speech in coarsely labeled audio and show both bag-level and instance-level results. We also study different pooling methods and how they can be adapted to densely distributed events as observed in our application. Finally, we show improvements using speech activity detection embeddings as features for foreground detection.

Original languageEnglish (US)
Article number7
JournalEurasip Journal on Audio, Speech, and Music Processing
Volume2021
Issue number1
DOIs
StatePublished - Dec 2021

Keywords

  • Foreground speech detection
  • Multiple instance learning
  • Weakly labeled audio
  • Wearable audio

ASJC Scopus subject areas

  • Acoustics and Ultrasonics
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Deep multiple instance learning for foreground speech localization in ambient audio from wearable devices'. Together they form a unique fingerprint.

Cite this