Skip to content

Short-time, Wavelet-inspired Mouse Submovement Detection

Home / Papers / Short-time, Wavelet-inspired Mouse Submovement Detection

Short-time, Wavelet-inspired Mouse Submovement Detection

Authors: Auejin Ham, Ben Boudaoud Date: 2026-04-22 Paper ID: openalex:2604.20673

Summary

This paper addresses the difficulty of decomposing overlapping ballistic submovements within human motor interaction data. The authors propose a wavelet-inspired segmentation technique specifically designed to identify these components from 1D speed time series. To enhance fitting accuracy, the method incorporates a self-weighted loss refinement step, effectively addressing limitations inherent in simpler wavelet-based approaches. Evaluation against standard segmentation baselines confirms the utility of this method for processing complex, high-frequency motor data.

Key Contributions

  • Proposes a wavelet-inspired method for detecting and parameterizing submovements in 1D speed time series.
  • Introduces a self-weighted loss refinement mechanism to mitigate fitting errors common in standard wavelet-based motion segmentation.
  • Validates the approach against dual-threshold and 1D persistence segmentation techniques using synthetic egocentric movement data with known ground truth.

Open Questions & Future Work

Archivist Review

I have approved the open question regarding robust submovement decomposition because it represents a clear, long-standing bottleneck in motor control analysis and human-computer interaction research. The proposed concepts were rejected because they are either minor implementation variations or too highly specialized to the domain of mouse-tracking to be considered reusable primitives in the broader time-series literature.

Approved Open Questions

  • Robust Submovement Decomposition Optimization: Improving submovement decomposition is critical for accurately characterizing human motor control strategies. Existing threshold-based or iterative methods often fail to account for high degrees of overlap or result in spurious detections, limiting their reliability for fine-grained behavioral analysis.

Rejected Candidates

  • [concept] Wavelet-inspired Submovement Detection (wavelet-inspired-submovement-detection) - not_reusable: The method is too domain-specific and narrow to warrant a general-purpose concept note, functioning more as an application of existing wavelet transforms to motion tracking.
  • [concept] Self-weighted Loss Refinement (self-weighted-loss-refinement) - subcomponent_of_broader_mechanism: This is a standard implementation detail for loss functions rather than a novel, reusable algorithmic concept.

Metadata & Links

url
https://arxiv.org/abs/2604.20673
paper_id
2604.20673
paper_source
openalex
domain
time-series
tags
architectures
datasets
concept_slugs
dataset_slugs
skill
TimeSeriesSkill
created_at
2026-04-25T05:38:30Z