MULASC is a multi-label database created as a contribution for the elaboration of the Freesound Database [1]. The deataset may be used for the development of experiments on multi-label acoustic scene classification. It contains 1,571 audio samples assigned to 34 different labels. It was created as part of a scientific initiation project, supported by Araucária Foundation, a Brazilian agency from the state of Paraná.
The database was developed by Viviane Shiraishi Besson, under the supervision of Professor Yandre Maldonado e Gomes da Costa, from State University of Maringá (UEM), in the state of Paraná, Brazil.
The samples were provided by the site Freesound (see full list here).
The characteristics of the database are summarized by the follow table.
This website provides the audio content of the database, spectrograms, visual features (extracted using Local Binary Pattern (LBP) which captures the textural content of the spectrogram images), and feature vectors labeled according to the Mulan framework format.
Content files
You can download the MULASC dataset here.
The rar file consists of a folder with audio samples in mono e wav format – the conversion was done using the software SoundConverter – and a pdf file with the samples information such as labels and credit.
Spectrograms
Spectrogram is a visual representation of the spectrum of the frequency of sounds. Considering that texture is the main visual attribute of the spectrograms, one can find some texture operators already successfully used to capture their content, like Local Binary Pattern (LBP), and Local Phase Quantization (LPQ). The Sound eXchange (SoX) software was used to generate the spectrograms starting from the audio signal.
They can be downloaded here.
Visual features
The LBP(8,2) texture features extracted from the audio clips, is available in here.
Mulan framework
Mulan is an open-source Java library for learning from multi-label datasets. The framework requires two files for the specification of a multi-label dataset [4]:
Both files can be downloaded here.
[1] Fonseca, E., Pons J., Favory X., Font F., Bogdanov D., Ferraro A., Oramas S., Porter A., & Serra X. Freesound Datasets: A Platform for the Creation of Open Audio Datasets. 18th International Society for Music Information Retrieval Conference, 2017.
[2] Tsoumakas, G., Katakis, I., Vlahavas, I. Mining Multi-label Data. In: Data Mining and Knowledge Discovery Handbook, O. Maimon, L. Rokach (Ed.), Springer, 2nd edition, 2010.