Dynamic sound source detection using partial whitening and a coherent steered response power algorithm with adaptive CFAR thresholding. 28 element irregular microphone array locates voice sounds every 20 ms. Detections are rendered in above stem plots corresponding to video recording. False positive threshold set to yield 1.5 false detections per frame. Dot on 3-D grid represents location, and its size is proportional to excess threshold crossing.

Loading more stuff…

Hmm…it looks like things are taking a while to load. Try again?

Loading videos…