The Pyramic Dataset contains recordings done using the Pyramic 48 channel microphone array in an anechoic chamber. The recordings consist of 8 different samples (2x sweeps, 1x noise, 5x speech) repeated at 180 angles (every 2 degrees) and from 3 different heights. The audio samples recorded are Linear and exponential sweeps Noise sequence 2x male and 3x female speech This dataset allows to evaluate the performance of array processing algorithms on real-life recordings done using MEMS microphones similar to those used in mobile phones with all the non-idealities involved. The dataset is suitable for both 2D and 3D scenarios. By subsampling the 48 microphones, a large number of array configurations can be tested. Example of algorithms are: Direction of arrival (DOA) estimation Beamforming Source separation Array calibration Another application is the generation of realistic room impulse by combining the impulse responses of microphones from sources at multiple angles with a variant of the image source model. In addition to the raw (compressed or not) and segmented recordings, the impulse responses of all the microphones for every source locations were recovered from the exponential sweep measurements and are distributed together with the dataset. The initial manual measurement of loudspeakers and microphones locations was improved upon using a blind calibration method. This record contains The compressed recordings (TTA format) Segmented recorded samples Impulse responses Documentation and code (also available on github) The raw measurements in wav format are available as a separate record (10.5281/zenodo.1209005). The best way to get started is to only get the documentation and code from github (a copy is available in pyramic-dataset-doc-d2a456b4.zip) and follow the instructions in the README. The version on github is most up-to-date. If possible, please use that one.

The author would like to acknowledge Juan Azcarreta Ortiz, Corentin Ferry, and René Beuchat for their help in the design and usage of the Pyramic array. Hanjie Pan, Miranda Kreković, Mihailo Kolundzija, and Dalia El Badawy for lending a hand, or even two, during experiments. Finally, Juan Azcarreta Ortiz, Eric Bezzam, Hanjie Pan and Ivan Dokmanić for feedback on the documentation and dataset organization.