Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

WildDESED: An LLM-Powered Dataset for Wild Domestic Environment Sound Event Detection System

Yang Xiao, Rohan Kumar Das

TL;DR

This work presents a new large language model (LLM)-powered dataset namely wild domestic environment sound event detection (WildDESED), crafted as an extension to the original DESED dataset to reflect diverse acoustic variability and complex noises in home settings.

Abstract

This work aims to advance sound event detection (SED) research by presenting a new large language model (LLM)-powered dataset namely wild domestic environment sound event detection (WildDESED). It is crafted as an extension to the original DESED dataset to reflect diverse acoustic variability and complex noises in home settings. We leveraged LLMs to generate eight different domestic scenarios based on target sound categories of the DESED dataset. Then we enriched the scenarios with a carefully tailored mixture of noises selected from AudioSet and ensured no overlap with target sound. We consider widely popular convolutional neural recurrent network to study WildDESED dataset, which depicts its challenging nature. We then apply curriculum learning by gradually increasing noise complexity to enhance the model's generalization capabilities across various noise levels. Our results with this approach show improvements within the noisy environment, validating the effectiveness on the WildDESED dataset promoting noise-robust SED advancements.

WildDESED: An LLM-Powered Dataset for Wild Domestic Environment Sound Event Detection System

TL;DR

This work presents a new large language model (LLM)-powered dataset namely wild domestic environment sound event detection (WildDESED), crafted as an extension to the original DESED dataset to reflect diverse acoustic variability and complex noises in home settings.

Abstract

This work aims to advance sound event detection (SED) research by presenting a new large language model (LLM)-powered dataset namely wild domestic environment sound event detection (WildDESED). It is crafted as an extension to the original DESED dataset to reflect diverse acoustic variability and complex noises in home settings. We leveraged LLMs to generate eight different domestic scenarios based on target sound categories of the DESED dataset. Then we enriched the scenarios with a carefully tailored mixture of noises selected from AudioSet and ensured no overlap with target sound. We consider widely popular convolutional neural recurrent network to study WildDESED dataset, which depicts its challenging nature. We then apply curriculum learning by gradually increasing noise complexity to enhance the model's generalization capabilities across various noise levels. Our results with this approach show improvements within the noisy environment, validating the effectiveness on the WildDESED dataset promoting noise-robust SED advancements.
Paper Structure (12 sections, 1 equation, 4 figures, 2 tables)

This paper contains 12 sections, 1 equation, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. WildDESED
  4. What type of background noises do we use?
  5. What are the domestic scenarios we choose?
  6. How do we mix the background noises to the scenarios?
  7. Curriculum Learning for Noise-robust SED
  8. Experimental Settings
  9. Dataset and Evaluation Metric
  10. Implementation Details
  11. Results and Discussion
  12. Conclusion

Figures (4)

  • Figure 1: Illustrationof Morning Routine Scenario out of the total eight scenarios in WildDESED dataset. In the scenario, key target sound events are written in bold fonts, along with added different background noises to simulate real-life settings.
  • Figure 2: Quadrant showing four groups of noise types based on their acoustic characteristics considered in the WildDESED.
  • Figure 3: Statistics of noises in the WildDESED subsets.
  • Figure 4: Statistics of the scenarios in the WildDESED subsets.