Face recognition on embedded devices: a proof of concept for residential application

Abstract

Face recognition is a well-known technique with a wide range of existing real world applications. Residential systems, like video intercom or security alarm, are instead nearly unfamiliar terrain to such methods with few commercial solutions available today in the market. A leading company in the sector has called for a research, with the purpose of assessing the feasibility of equipping their embedded video intercom systems with a feature for automatic identification and authorization of the calling subject. In this paper a combination of face detection and recognition methods on such system has been studied and evaluated, leading to a proof of concept in order to verify the feasibility assumptions, and support the claimant company decision process on investing for prototype development and final product extensions and adjustments. Early promising results suggest that the proposed system could prove usable, complying with constraints such as providing a reasonable recognition rate and execution time, running on embedded hardware and being user-friendly.