Temporal noise acts like sensor bias in wearable activity sensors
When Temporal Perturbations Act Like Sensor Biases: Label-Free Auditing of Wearable Activity Recognizers
Machine LearningArtificial IntelligenceCryptography and Security
Summary
Wearable devices that track human activities can sometimes mistake steady sensor errors for actual motion changes. The authors present SpectrumAudit, a method that tests these devices without needing activity labels to see if constant sensor offsets harm accuracy. They found that for many devices and datasets, these steady errors cause bigger mistakes than normal fluctuations. This helps identify when sensor biases are the real problem behind misclassifications.
What this means in practice
- •For wearable device engineers: Assess if persistent sensor offsets rather than dynamic noise cause activity recognition errors without using labeled data from the sensor users.
- •For mobile health app developers: Improve app reliability by identifying and mitigating accuracy losses due to sensor biases during user calibration phases.
Authors
Qingyu Wu, Yuan Wei, Renju Liu, Hua Cheng
Abstract
Wearable human-activity recognition (HAR) models operate across sensors, subjects, and backbones, yet a smooth waveform may appear temporal while exploiting a persistent sensor offset primarily. We introduce SpectrumAudit, a label-sealed audit that fits a phase-randomized full-window stimulus on calibration windows from subjects held out from training and testing. After selection, it replays its exact DC projection and budget-constrained zero-mean residual on the same frozen victim without refitting. Across 27 victims from three datasets and three backbones, the selected waveforms cause 2.87-40.83-point three-phase robust accuracy losses. Under this replay budget, DC is more damaging than AC on 24/27 victims and recovers at least 90% of the full drop on 22/27; all 5 failures occur on WISDM. In a held-out UTD-MHAD check, the selected waveform causes 13.49-pp accuracy and 11.68-pp macro-F1 losses, versus -0.66 pp for matched random changes. The audit diagnoses offset versus zero-mean variation under a common peak-budget cap. The code will be released upon acceptance.