Actions are recognized by posture, not by trajectory express
Researchers took 16 everyday activities — from washing dishes to squatting — and broke the movements down into elementary components, much like a spectroscope splits light into a spectrum. Comparing three approaches, they found an unexpected key to recognition: a single averaged snapshot is enough. Only nine joints out of dozens carried the crucial information — the rest hardly mattered.
Paradoxically, a static posture, similar to the overall brightness of a star in photometry, turned out to be more important than dynamics. The brain recognizes an action instantly, without scrutinizing the process. But for reproduction, a full timeline is needed: one method reconstructed a natural gait, the other only a frozen statue.
This principle of separation formed the basis of a standard model for rapid clinical tests: a single snapshot can flag impairments, while rehabilitation will require detailed video recording. That’s how the brain saves resources, and medicine gets a simple diagnostic tool.
Original abstract
as written by the authors arXiv:2607.13216
Humans recognize movements effortlessly, even from noisy and complex visual input. But what information in the stimulus allows humans to rapidly classify movements? No framework has systematically compared different strategies of movement analysis to address this question. Here, we used videos of 16 daily activities from the MoVi dataset and compared three strategies: Temporal Movement Primitives (TMPs), which decompose movements into weighted sums of temporally smooth basis functions; Legendre polynomial coefficients, which project joint-coordinate trajectories onto an orthogonal polynomial basis; and Autoencoder latent embeddings. Legendre coefficients and TMPs achieved the highest classifier accuracy, followed by autoencoders. We found two discriminative features for movement classification. The most informative is the general posture of the body, the average spatial configuration that distinguishes one activity from another. Additionally, we identified 9 critical joints that are most predictive for movement classification. Interestingly, good classification accuracy did not automatically lead to good movement generation: when we reconstructed movements for each activity, TMPs preserved the temporal dynamics and produced perceptually natural motion, whereas reconstructions from Legendre coefficients retained only the average posture and appeared frozen. These results reveal a dissociation in how movement information is organized: the static configuration of the body suffices to classify what activity is performed, but the temporal dynamics of movement are required to reconstruct how it unfolds. This distinction clarifies which features the visual system may rely upon for rapid action recognition, and suggests that postural features could enable efficient movement screening in clinical applications, while dynamic information remain essential wherever movement generation is the goal.
Out of all the body's joints, only nine proved crucial for recognizing any movement.