Performance of an Active THz Imaging System for Recognition of Concealed Faces


Journal of Infrared, Millimeter, and Terahertz Waves, vol.44, no.5-6, pp.365-378, 2023 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 44 Issue: 5-6
  • Publication Date: 2023
  • Doi Number: 10.1007/s10762-023-00925-2
  • Journal Name: Journal of Infrared, Millimeter, and Terahertz Waves
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aerospace Database, Communication Abstracts, Compendex, Metadex, Civil Engineering Abstracts
  • Page Numbers: pp.365-378
  • Keywords: Biometrics, Face recognition, Terahertz imaging
  • Middle East Technical University Affiliated: Yes


Millimeter wave (MMW) and submillimeter wave (SMW) imaging techniques have a wide range of security applications due to their characteristics such as transmission through barrier materials. Biometric recognition based on MMW and SMW imaging has been attracting attention as an alternative to recognition methods based on imaging human body in visible band, since these systems pose a number of limitations such as being sensitive to illumination conditions and occlusions. This study explores the possibility of biometric recognition based on terahertz (THz) face images. The THz images are acquired by an active imaging system which utilizes a transceiver operating at 340 GHz in frequency modulated continuous wave (FMCW) mode. We examine if these images contain sufficient information on facial characteristics for distinguishing individuals, using a similarity comparison between images of different individuals. Analysis is based on a set of THz face images of 20 individuals. Images of the head areas of the participants have been recorded with and without a concealing clothing, in order to explore the possibility of identifying individuals with concealed faces. Analysis within this closed data set indicates that imaging at 340 GHz has sensitivity to the facial characteristics which are preserved in images through clothing, and can be used for biometric recognition.