audioTextureDatasets

This repository has two functions;

1. To act as a database for audio texture datasets with and without control parameters, for use in machine learning models.
2. To provide useful functionalities for manipulating audio texture datasets for machine learning models.

A repository holding a collection of audio texture datasets for training machine learning model, with a stadardised approach.

Datasets

Audio Texture Datasets

• UrbanSound8K, https://urbansounddataset.weebly.com/urbansound8k.html
• BBC Sound Effects, https://sound-effects.bbcrewind.co.uk/
• FreeSound, https://freesound.org/
• ESC-50, https://github.com/karolpiczak/ESC-50
• DCASE, https://dcase.community/challenge2023/task-foley-sound-synthesis

Audio Texture Datasets with control parameters

• SynTex, https://syntex.sonicthings.org/
• Parameterized Sound Set database, https://sonicthings.org:9999/
• Boreilly dataset, https://animatedsound.com/datasets/

Functionalities

Curerntly scripts exist for loading data and manipulating in the following ways;

File formatting

Scripts exist for manipulating audio files in a number of ways;

Audio length

Any length of audio can be split into any arbitrary length

Sample Rate

Audio can be re-sampled to;

• 44kHz
• 22kHz
• 16kHz
• 8kHz

Bit depth

Bit depth of Audio clips can be changed to;

• 16 bit
• 32 bit

Audio representations

Scripts exist for manipulating audio files in the time domain into a number of different representations for machine learning;

• Granulating waveform
• Spectrograms
• Mel Frequency Spectrograms
• Cepstral Coefficients
• Mel Freq Cepstral Coefficients