Sensor-Specific Transformer (PatchTST) Ensembles with Test-Matched Augmentation

TL;DR

This paper introduces a sensor-specific ensemble approach using PatchTST transformers with test-matched augmentation to improve robustness in human activity recognition under noisy sensor conditions.

Contribution

It proposes a novel ensemble method with sensor-specific models trained on augmented data to handle real-world noise in HAR tasks.

Findings

Achieved significantly higher macro-F1 score than baseline.

Demonstrated effectiveness of test-matched augmentation for noise robustness.

Validated approach on the 2nd WEAR Dataset Challenge.

Abstract

We present a noise-aware, sensor-specific ensemble approach for robust human activity recognition on the 2nd WEAR Dataset Challenge. Our method leverages the PatchTST transformer architecture, training four independent models-one per inertial sensor location-on a tampered training set whose 1-second sliding windows are augmented to mimic the test-time noise. By aligning the train and test data schemas (JSON-encoded 50-sample windows) and applying randomized jitter, scaling, rotation, and channel dropout, each PatchTST model learns to generalize across real-world sensor perturbations. At inference, we compute softmax probabilities from all four sensor models on the Kaggle test set and average them to produce final labels. On the private leaderboard, this pipeline achieves a macro-F1 substantially above the baseline, demonstrating that test-matched augmentation combined with transformer-based ensembling is an effective strategy for robust HAR under noisy conditions.