Soft biometrics: a survey

Abstract The field of biometrics research encompasses the need to associate an identity individual based on persons physiological or behaviour traits. While use intrusive techniques such as retina scans and finger print identification has resulted in highly accurate systems, scalability systems real-world applications surveillance border security been limited. As a branch research, origin soft could be traced back for non-intrusive solutions extracting traits person. Following high number outcomes reported literature biometrics, this paper aims consolidate scope across four thematic schemes (i) detailed review data sets, their annotation strategies building largest novel collection traits; (ii) assessment metrics that affect performance system; (iii) comparative analysis feature modality level fusion enhancing system performance; (iv) hybrid recognition using multi-scale criterion. also presents global associated person gender, age ethnicity. contribution is provide comprehensive scientific literature, identify open challenges offer insights new directions filed.