Audio-visual twins database

Abstract

Identical twins pose an interesting challenge for recognition systems due to their similar appearance. Although various biometrics have been proposed for the problem, existing works are quite limited due to the difficulty of obtaining a twins database. To encourage the methods for twins recognition and make a fair comparison of them by using the same database, we collected an audio-visual twins database at the Sixth Mojiang International Twins Festival held on 1 May 2010, China. Our database contains 39 pairs of twins in total, including Chinese, American and Russian subjects. This database contains several face images, facial motion videos and audio records for each subject. In this paper, we describe the collection procedure, organization of the database, and usage method of the database. We also show our experiments on face verification, facial motion verification and speaker verification for twins to provide usage examples of the database.