OPECT motion recognition in a real scene. (a) Schematic representation of extraction of keyframes from the video. (b) Temporal parameters and selection criteria for optimal AA concentrations in dynamic information filtration. (c) DA-mediated ANN architecture in image recognition. (d) Dynamic information retention profiles processed with varying AA concentrations across different temporal scales (t_d=0.5–2 s; t_s=3–16 s). (e) Recognition accuracy metrics with varying DA concentrations derived from ANN-based analysis of filtration outcomes.