Blindly counting the number of speech sources (talkers) in a meeting room can be a difficult task. This paper was presented at HSCMA 2017 at the Google Offices in San Francisco and shows how using coherent-to-diffuse ratios could allow real-time source counting.
