In pediatric healthcare, the accurate prediction of movement symmetry in newborns is crucial as a foundational component for future automated screening systems. The long-term goal is to develop anomaly detection tools that can identify deviations from normal movement patterns by comparing observed data with model predictions. However, achieving this objective first requires robust forecasting models capable of accurately capturing the complex dynamics of typical infant movement. Traditional time series forecasting methods often struggle to capture the complex, non-linear dependencies present in real-world temporal data, particularly in the context of infantile movement patterns. To address these challenges, we introduce the Signature Enhanced Decomposition Transformer (SEDformer), which integrates path signatures from rough path theory into the FEDformer architecture by replacing Fourier transforms with signature-based operations. Our experiments demonstrate that SEDformer achieves superior forecasting performance compared to FEDformer on a specialized dataset of infantile movement patterns. While the current study focuses on prediction accuracy evaluation against this baseline model, these results establish the foundation for future anomaly detection applications in early screening of motor development issues.