Jadwal Sholat

Memuat jadwal sholat…

Pertanian & Lingkungan editorial

Open AccessOA2023

Analysis and Acoustic Event Classification of Environmental Data Collected in a Citizen Science Project

This study analyzes acoustic data from the Sons al Balcó citizen science project, comparing the 2020 and 2021 campaigns and training a convolutional neural network to detect and classify sound events, achieving over 50% macro -score for prevalent noise sources.
Daniel Bonet-Solà; E. Vidaña-Vila; R. Alsina-Pagès· International Journal of Environmental Research and Public Health· 2023· DOI 10.3390/ijerph20043683

The core problem

The Sons al Balcó project leverages citizen science to study the soundscape in Catalonia, particularly during and after the COVID-19 lockdown. The project aims to gather acoustic data through citizen contributions and develop tools for automatic sound event detection to assess soundscape quality. This paper focuses on the analysis of data from two collection campaigns (2020 and 2021) and the application of a convolutional neural network (CNN) for classifying acoustic events, even when they occur simultaneously. The motivation stems from the need to process large-scale citizen-collected data efficiently and to understand the impact of the pandemic on environmental noise.

Innovation

The project conducted two campaigns: in 2020, 365 videos were collected, and in 2021, 237 videos were obtained. The acoustic samples were analyzed and compared. For automatic event detection, a convolutional neural network was trained to classify sound events from the audio recordings. The network was designed to handle polyphonic soundscapes where multiple events can occur simultaneously. The performance was evaluated using event-based macro -score, focusing on the most prevalent noise sources. The methodology includes data preprocessing, network architecture design, training, and evaluation. The CNN likely uses spectrogram representations of audio as input, with convolutional layers for feature extraction and fully connected layers for classification.
Introduction
The Sons al Balcó project leverages citizen science to study the soundscape in Catalonia, particularly during and after the COVID-19 lockdown. The project aims to gather acoustic data through citizen contributions and develop tools for automatic sound event detection to assess soundscape quality. This paper focuses on the analysis of data from two collection campaigns (2020 and 2021) and the application of a convolutional neural network (CNN) for classifying acoustic events, even when they occur simultaneously. The motivation stems from the need to process large-scale citizen-collected data efficiently and to understand the impact of the pandemic on environmental noise.

Why it matters

The study demonstrates the feasibility of using citizen science data and machine learning for soundscape analysis. The CNN's performance, while moderate, highlights the challenges of polyphonic acoustic event detection. The influence of prevalence and FBR suggests that class imbalance and signal-to-noise ratio are critical factors. The authors discuss the limitations of the current approach and potential improvements, such as data augmentation, advanced architectures, or transfer learning. The findings also provide insights into the soundscape of Catalonia during the pandemic, with potential implications for noise pollution monitoring and urban planning. Future work could involve expanding the dataset and refining the model for broader categories.

Who should read this

Pembaca Research Digest dan praktisi rumpun ini.

Opening member content…