The Red Hen Anonymizer and the Red Hen Protocol for de-identifying audiovisual recordings

Abstract

Scientists of multimodal communication have no established policy or default tool for sharing deidentified audiovisual recordings. Recently, new technology has been developed that enables researchers to deidentify voice and appearance. These software tools can produce output in JSON format that specifies bodypose and face and hand keypoints in numerical form, suitable for computer search, machine learning, and sharing. The Red Hen Anonymizer is a new tool for de-identification. This article presents the Red Hen Anonymizer and discusses guidelines for its use.