Isolating Fast and Slow Flows in Three-Dimensional Fluid Dynamics

Abstract Isolating and analyzing different velocity components within fluid flows is fundamental to investigation in several scientific disciplines. Current optical flow (OF) analyses fail when tracking fast flows in the presence of superimposed slower ones due to inherent small-motion assumptions. We present an approach that overcomes this limitation using two techniques: intensity rebalancing and per-pixel time-based high-pass filtering. Using speech airflow visualization as a demonstra- tion, we isolated rapid speech-generated airflows from slower buoyancy-driven thermal flows—a distinction impossible to achieve with conventional OF. Our method boosted explanatory power (R2:2.64%→7.57%), reduced model com- plexity, and improved fit (fREML: 6.8E7→7.3E6 = 89.3% residual reduction), revealing distant airflow patterns that explain known speech airflow influences on speech perception[1–3]. This approach requires no tracer particles or contrast agents and is generalizable to analysis of any three-dimensional semi-transparent flows with relevant trajectories along the longitudinal plane that contain simultaneous high and low speed components.