Humans can track but fail to predict accelerating objects

Abstract

Objects in our visual environment often move unpredictably and can suddenly speed up or slow down. The ability to account for acceleration when interacting with moving objects can be critical for survival. Here, we investigate how human observers track an accelerating target with their eyes and predict its time of reappearance after a temporal occlusion by making an interceptive hand movement. Before occlusion, the target was initially visible and accelerated for a brief period. We tested how observers integrated target motion information by comparing three alternative models that predicted time-to-contact (TTC) based on the (1) final target velocity sample before occlusion, (2) average target velocity before occlusion, or (3) target acceleration. We show that visually-guided smooth pursuit eye movements reliably reflect target acceleration prior to occlusion. However, systematic saccade and manual interception timing errors reveal an inability to consider acceleration when predicting TTC. Interception timing is best described by the final velocity model that relies on extrapolating the last available velocity sample before occlusion. These findings provide compelling evidence for differential acceleration integration mechanisms in vision-guided eye movements and prediction-guided interception and a mechanistic explanation for the function and failure of interactions with accelerating objects.

Texture and Flag Color Extraction in Backpropagation Neural Network Architecture

“…The use of computers as human aids is highly expected to the extent that computer capabilities can replace the limitations possessed by humans (Trunk et al, 2020;Korteling et al, 2021). Humans can recognize an object using their eyes and brain, but if they cannot work well, it will hinder human work (Sun et al, 2020;Kreyenmeier et al, 2021). Artificial Neural Network (ANN) or Artificial Neural Network (ANN) is an information processing system with characteristics resembling the human nervous system which can solve SVM and KNN problems by training on large data and ANN has the ability to tolerate errors so it can produce predictions good (Putra & Ulfa Walmi, 2020;Nofitriyani et al, 2022).…”