CatMeows: A Publicly-Available Dataset of Cat Vocalizations

Abstract

This dataset, composed of 440 sounds, contains meows emitted by cats in different contexts. Specifically, 21 cats belonging to 2 breeds (Maine Coon and European Shorthair) have been repeatedly exposed to three different stimuli that were expected to induce the emission of meows:

1. Brushing - Cats were brushed by their owners in their home environment for a maximum of 5 minutes;
2. Isolation in an unfamiliar environment - Cats were transferred by their owners into an unfamiliar environment (e.g., a room in a different apartment or an office). Distance was minimized and the usual transportation routine was adopted so as to avoid discomfort to animals. The journey lasted less than 30 minutes and cats were allowed 30 minutes with their owners to recover from transportation, before being isolated in the unfamiliar environment, where they stayed alone for maximum 5 minutes;
3. Waiting for food - The owner started the routine operations that preceded food delivery in the usual environment the cat was familiar with. Food was given at most 5 minutes after the beginning of the experiment.

The dataset has been produced and employed in the context of an interdepartmental project of the University of Milan (for further information, please refer to this doi). The content of the dataset has been described in detail in a scientific work currently under review; the reference will be provided as soon as the paper is published.

File naming conventions 

Files containing meows are in the dataset.zip archive. They are PCM streams (.wav).
Naming conventions follow the pattern C_NNNNN_BB_SS_OOOOO_RXX, which has to be exploded as follows:

• C = emission context (values: B = brushing; F = waiting for food; I: isolation in an unfamiliar environment);
• NNNNN = cat’s unique ID;
• BB = breed (values: MC = Maine Coon; EU: European Shorthair);
• SS = sex (values: FI = female, intact; FN: female, neutered; MI: male, intact; MN: male, neutered);
• OOOOO = cat owner’s unique ID;
• R = recording session (values: 1, 2 or 3)
• XX = vocalization counter (values: 01..99)

Extra content

The extra.zip archive contains excluded recordings (sounds other than meows emitted by cats) and uncut sequences of close vocalizations.

Terms of use

The dataset is open access for scientific research and non-commercial purposes.

The authors require to acknowledge their work and, in case of scientific publication, to cite the most suitable reference among the following entries:

Ntalampiras, S., Ludovico, L.A., Presti, G., Prato Previde, E., Battini, M., Cannas, S., Palestrini, C., Mattiello, S.: Automatic Classification of Cat Vocalizations Emitted in Different Contexts. Animals, vol. 9(8), pp. 543.1–543.14. MDPI (2019).
ISSN: 2076-2615

Ludovico, L.A., Ntalampiras, S., Presti, G., Cannas, S., Battini, M., Mattiello, S.: CatMeows: A Publicly-Available Dataset of Cat Vocalizations. In: Li, X., Lokoč, J., Mezaris, V., Patras, I., Schoeffmann, K., Skopal, T., Vrochidis, S. (eds.) MultiMedia Modeling. 27th International Conference, MMM 2021, Prague, Czech Republic, June 22–24, 2021, Proceedings, Part II, LNCS, vol. 12573, pp. 230–243. Springer International Publishing, Cham (2021). 
ISBN: 978-3-030-67834-0 (print), 978-3-030-67835-7 (online)
ISSN: 0302-9743 (print), 1611-3349 (online)