T2M Mamba: Motion Periodicity-Saliency Coupling Approach for Stable Text-Driven Motion Generation
T2M Mamba addresses two issues in text-to-motion generation: ignoring how motion periodicity and keyframe importance interact, and instability with paraphrases. The framework uses Periodicity-Saliency Aware Mamba to couple these factors via DPC for keyframe weighting and FFT for periodicity estimation. It also introduces a Periodic Differential Cross-modal Alignment Module (PDCAM) to ensure stable alignment between text prompts and motion sequences.