مقاله انگلیسی رایگان در مورد یک روش اصلاح شده برای تشخیص چهره در شرایط تخریب شده – الزویر 2018

Interest in using individual differences in face recognition ability to better understand the perceptual and cognitive mechanisms supporting face processing has grown substantially in recent years. The goal of this study was to determine how varying levels of face recognition ability are linked to changes in visual information extraction strategies in an identity recognition task. To address this question, fifty participants completed six tasks measuring face and object processing abilities. Using the Bubbles method (Gosselin & Schyns, 2001), we also measured each individual’s use of visual information in face recognition. At the group level, our results replicate previous findings demonstrating the importance of the eye region for face identification. More importantly, we show that face processing ability is related to a systematic increase in the use of the eye area, especially the left eye from the observer’s perspective. Indeed, our results suggest that the use of this region accounts for approximately 20% of the variance in face processing ability. These results support the idea that individual differences in face processing are at least partially related to the perceptual extraction strategy used during face identification.

Introduction

Face identification is a great challenge for the visual system, as human faces consist of a small set of facial features (e.g. the eyes, the nose, the mouth) with only subtle variations in inter-attribute distances (Dupuis-Roy, Fiset, Dufresne, Caplette, & Gosselin, 2014; TaschereauDumouchel, Rossion, Schyns, & Gosselin, 2010; see also Burton, Schweinberger, Jenkins, & Kaufmann 2015; Sandford & Burton, 2014). In the last few decades, the processes supporting face identification have been extensively investigated using group-based approaches where interindividual variations were typically regarded as uninformative noise. However, significant variations in face identification ability have been observed within the healthy population (Bate, Parris, Haslam, & Kay, 2010; Bowles et al., 2009; Duchaine & Nakayama, 2006; Royer, Blais, Gosselin, Duncan, & Fiset, 2015; Wilmer et al., 2010), and many authors now highlight the importance of individual differences to gain a better understanding of face processing mechanisms (e.g. Yovel, Wilmer, & Duchaine, 2014; see also Richler, Cheung, & Gauthier, 2011 for a discussion). An example of this growing interest for individual differences is found in recent papers studying holistic processing, i.e. the extent to which individuals integrate facial parts into a unified whole or “gestalt” (Farah, Wilson, Drain, & Tanaka, 1998; see Richler, Palmeri, & Gauthier, 2012 for precisions regarding the measures and subtypes of holistic processing). The experimental effects thought to measure holistic processing (e.g. composite effect, Young, Hellawell, & Hay, 1987; part-whole task, Tanaka & Farah, 1993) have been replicated numerous times at the group-average level (see Richler et al., 2012). However, if holistic processing is indeed important for face processing and identification, individual differences in the ability to discriminate and recognize faces might be expected to at least partly depend on this mechanism. Results addressing this question are mixed: While some have obtained a significant correlation between face recognition ability and the magnitude of certain experimental effects thought to reflect holistic processing (DeGutis, Wilmer, Mercado, & Cohan, 2013; Richler et al., 2011; Wang, Li, Fang, Tian, & Liu, 2012), others have not (Konar, Bennett, & Sekuler, 2010; Richler, Floyd, & Gauthier, 2014). Moreover, studies finding a link indicate differences in holistic face perception only account for a limited proportion of differences in face recognition ability. We thus believe it is important to investigate other perceptual and cognitive mechanisms known to be involved, on average, in face recognition using an individual differences based approach. Here, we explore the hypothesis that the visual information extracted during face recognition is systematically related to face processing abilities. In line with this proposition, Pachai, Sekuler, & Bennett (2013) demonstrated that tuning for horizontal information is significantly correlated with upright face identification accuracy as measured within the same recognition task (see also Pachai, Sekuler, Bennett, Schyns, & Ramon, 2017).