MELAUDIS: A Large-Scale Benchmark Acoustic Dataset For Intelligent Transportation Systems Research
View abstract on PubMed
Summary
This summary is machine-generated.Researchers developed MELAUDIS, a large real-world acoustic dataset for intelligent transportation systems (ITS). This dataset enhances vehicle detection and classification, improving accuracy with advanced audio processing techniques.
Area Of Science
- Intelligent Transportation Systems (ITS)
- Acoustic Signal Processing
- Machine Learning for Transportation
Background
- Acoustic traffic sensors offer cost-effective road traffic monitoring.
- Existing audio datasets for ITS lack real-world complexity and diversity.
- There is a need for comprehensive acoustic data to advance intelligent transportation research.
Purpose Of The Study
- Introduce MELAUDIS, the first comprehensive real-world acoustic dataset for ITS.
- Facilitate vehicle detection, traffic status monitoring, and vehicle type classification.
- Establish a benchmark dataset to drive innovation in urban acoustic intelligence.
Main Methods
- Collected diverse audio recordings from multi-lane, two-way traffic roads.
- Included various traffic conditions, ambient noises, and weather settings.
- Annotated six vehicle types (bicycles, motorcycles, cars, buses, trucks, trams) in single and multi-vehicle scenarios.
Main Results
- MELAUDIS comprises 5,792 background noise, 7,345 vehicle sound, and 2,955 idling sound recordings.
- Achieved a classification accuracy improvement from 65.1% to 82.84% using log-mel-spectrograms and Convolutional Neural Networks (CNNs).
- The dataset's extensive labeling (1,200+ man-hours) ensures high quality for research.
Conclusions
- MELAUDIS is the largest urban acoustic dataset to date, addressing critical data gaps in ITS.
- The dataset significantly improves the performance of acoustic-based vehicle analysis.
- MELAUDIS provides a robust foundation for future research and development in intelligent transportation systems.

