Omarrran/environmental_sound_data_hnm

Environmental Sound Collections Dataset

This dataset contains metadata about environmental sound recordings from various popular collections.

Contents

• environmental_sound_data/all_environmental_sounds_metadata.csv: Combined metadata for all sound clips (21680 records)
• collection_statistics.csv: Summary statistics for each collection
• category_distribution.csv: Distribution of sound categories across collections
• recording_device_statistics.csv: Statistics about recording devices used
• sound_quality_index.csv: Derived dataset with quality scores for each clip
• sound_catalog.json: Searchable catalog with sample clips and collection information

Collections Included

• UrbanSound8K: A dataset containing 8732 labeled sound excerpts (<=4s) of urban sounds from 10 classes. (8732 clips)
• ESC-50: ESC-50 is a labeled collection of 2000 environmental audio recordings suitable for benchmarking classification methods. (2000 clips)
• AudioSet: AudioSet consists of an expanding ontology of 632 audio event classes and a collection of 2M+ human-labeled 10-second sound clips. (2084320 clips)
• FSD50K: FSD50K is an open dataset of human-labeled sound events containing 51,197 audio clips totaling over 100 hours of audio. (51197 clips)
• TUT Acoustic Scenes: TUT Acoustic Scenes dataset consists of recordings from 15 different acoustic environments. (4680 clips)

Fields Description

• clip_id: Unique identifier for the sound clip
• collection: Source collection name
• category: Primary sound category
• subcategory: Secondary sound category (if applicable)
• duration_seconds: Length of the recording in seconds
• sample_rate_hz: Sampling rate in Hz
• bit_depth: Audio bit depth (16 or 24 bit)
• channels: Number of audio channels (1=mono, 2=stereo)
• file_format: Audio file format (wav, mp3, flac, ogg)
• file_size_kb: Estimated file size in kilobytes
• signal_to_noise_ratio: Signal-to-noise ratio (higher = cleaner recording)
• recording_device: Device used to record the sound
• location_type: Type of recording location (urban, nature, indoor, domestic)
• indoor: Whether the recording was made indoors (1) or outdoors (0)
• timestamp: Recording date
• license: License under which the sound is available

Usage Examples

This metadata can be used for:

1. Finding suitable environmental sounds for machine learning projects
2. Analyzing the composition of environmental sound datasets
3. Comparing quality and characteristics across different sound collections
4. Identifying distribution biases in audio collections

Data Generation Note

This dataset contains synthetic metadata that represents the structure and characteristics of actual environmental sound collections, but the specific clip details are generated for demonstration purposes.

Created: 2025-04-26

NOTE:

All the files are packaged into a ZIP archive (environmental_sound_collections_data.zip) for easy download. The dataset provides a comprehensive view of environmental sound collections while being compact enough to work with on standard hardware.