name: Rasheed Mudasiru
summary: A 63-minute recording featuring natural conversations among multiple speakers in a realistic acoustic environment, with noticeable background noise. The dataset was created for use in audio forensics, multi-speaker separation, and speech enhancement research as part of an undergraduate project.
category: Speech / Audio Forensics / Signal Processing
license: MIT License
file: audio-for-forensic-and-multi-speaker-separation.zip The corpus archive.