RAMDC: Room-Aware Multi-Device Clustering for Large Scale Teleconferencing Dataset

RAMDC-Dataset is a simulation-generated dataset for room-aware multi-device clustering in large-scale teleconferencing. In each scenario, multiple devices simultaneously play a short chirp. The dataset contains three multi-room settings with 4 rooms, 7 rooms, and 11 rooms, each with 50 synthesized scenarios. For every scenario we include (i) device chirp recordings, (ii) deconvolution-based aggregated responses per device, and (iii) energy-decay curves (EDCs) (broadband + octave bands). The mixtures are synthesized by convolving chirps with real room impulse responses (GTU-RIR: https://github.com/mehmetpekmezci/gtu-rir).