This study tackles two major challenges in automated detection of autism-related self-stimulatory behaviors: (1) identifying the optimal sequence-based neural network architecture and temporal sampling rate, and (2) characterizing data augmentation strategies for training on small behavioral datasets.
For the first objective, Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models were trained on pose-derived features from the Self-Stimulatory Behavior Diagnosis (SSBD) dataset at frame sampling intervals of 1, 5, 15, 30, 45, and 90 frames. Both architectures surpassed previous Convolutional Neural Network (CNN) baselines (62-76% accuracy), achieving peak accuracies of 97.5% (LSTM) and 98.75% (GRU) at a sampling interval of every 15 frames.
For the second objective, ten data augmentation strategies were applied to an I3D transfer learning pipeline, with an ablation study quantifying the marginal contribution of each technique. Horizontal flip achieved the highest standalone accuracy (48.78%), while the exclusion of upsampling from the augmentation pipeline caused the largest performance degradation, indicating its necessity for complex behavioral video augmentation. A personalized machine learning approach, where per-subject models were trained and tested on temporally split segments of each video, produced consistent predictions (mean loss 1.84, SD 0.79). These results provide practitioners with concrete guidance on architecture selection, sampling rate, and augmentation strategy for video-based behavioral classification in data-scarce clinical domains.
Blogger's Review: This study not only demonstrates the effectiveness of various neural network architectures in autism behavior classification but also emphasizes the crucial role of data augmentation strategies, particularly when dealing with small datasets, thereby showcasing significant clinical application potential.