This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility: Researchers led by Carina Hahn at the National Institute of Standards and Technology (NIST), U.S., have found that forensic specialists in facial identification need only one or two features, such as eyes or ears, to identify a face when looking for a match. Published in PLOS One, the study found that specialists find ears to be the most useful but are capable of adapting when recognizing that another feature might be more beneficial.
Forensic face specialists help investigations and provide expert testimony at court trials. Understanding the cognitive processes that they use when identifying faces will be useful for improving how future experts are trained, as well as for improving machine-learning algorithms that are used for facial recognition. In the study, face specialists were given photos of two similar faces side by side and a hidden face.
For each face pair, their job was to identify which of the two faces matched the hidden face, with as little information as possible. To accomplish this, they requested the hidden facial features to be revealed one by one. Accuracy averaged over 90% correct, and the experts were able to identify the face after seeing only one or two features.
Seeing the ears and forehead led to the highest accuracies in identification, with necks and chins being the least helpful. While ears were the most frequently self-selected feature, the other features were not selected based on their usefulness. This was particularly true of the forehead, which was not frequently selected but was associated with the highest accuracy.
When not choosing the ears, experts adapted as hypothesized and generally requested features based on how different that feature was between the two faces. The findings do not reflect real-world accuracy, as the conditions under which the study was conducted were artificial and designed to address certain research questions. The authors also note the facial features they analyzed may not generalize well to different sets of images, especially as the images primarily depicted white and East Asian individuals.
Nevertheless, these findings can be used to improve training of both humans and machine-learning models for facial identification. Hahn adds, "What surprised me most was that specialists could accurately judge identity from a single facial feature most of the time. It's remarkable how much identity information they can extract from so little." "Specialists have a striking ability to adapt their strategy depending on the face in front of them, rather than applying one fixed rule.
Extract — continue reading at the source.
