Continuous development in surveillance systems is increasingly motivating research in biometrics to articulate unconstrained recognition of human faces. Comparative soft biometric has recently been employed to characterize eyewitness testimonies for use in a biometric system to improve the recognition accuracies of the traditional biometric systems. In this paper, we present a face profile recognition system by fusing features extracted in a traditional face recognition system and eyewitness testimonies processed in a soft biometric system to improve recognition accuracies. Here we have also demonstrated an association between our traditional face profile biometric system and the soft biometric system by numerically mapping the features extracted from a face profile to its soft biometric attributes. Our experiments on 230 subjects in XM2VTSDB dataset demonstrate 84% accuracy for traditional biometrics. The recognition rate is further improved to 98% accuracy when soft and traditional biometrics are fused.