Privacy-preserving People Identification and Tracking without Use of Facial Recognition

Nowadays, facial recognition technology is widely being used for people identification and tracking in both consumer and industrial applications. Because of near-perfect accuracy of such systems, these are being actively deployed worldwide especially in public safety and retail industries. However, there are serious privacy and civil rights concerns due to the inherent nature of analyzing the human face. As a result, several major cities (e.g. San Francisco) have banned use of facial recognition technology by the police and other agencies. In addition, several major technology companies have also announced plans to either quit facial recognition business altogether or being selective in who they supply this product to. In this paper, we propose an alternative approach for people identification for use in commercial premises such as productivity measurement in a factory or retail. Our proposed system consists of a camera to locate and track people anonymously and a wireless beacon transmitter for an opt-in-based people identification that does not require facial/gait or appearance attributes. By intelligently fusing this multimodal data and leveraging recent advances in deep learning, the proposed system locates every person in frame and identifies those who opted-in for people identification by equipping them with a wireless beacon transmitter.