Human motion generation improves with frequency separated tokens
Rethinking Fixed Temporal Grids: Frequency-Disentangled Motion Generation
Computer Vision and Pattern Recognition
Summary
Current ways to create digital human movements break time into equal pieces, but real movements have a mix of slow and fast parts. This makes it hard to capture quick details like foot taps without blurring them. The authors propose a method called FreqMo that splits motion into parts by how quickly they change over time, keeping both slow moves and fast details clear. They use a new way to compress these parts together, making the system smaller and better at creating realistic motion.
What this means in practice
- •For animation studios: Generate more realistic human motion with clearer fast movements like footsteps and gestures in digital characters.
- •For video game developers: Produce smoother and more detailed character animations that capture both slow and rapid motions without extra computational cost.
Authors
Yunjiao Zhou, Junlang Qian, Gen Li, Xinying Guo, Lihua Xie, Jianfei Yang
Abstract
Most human motion generation methods encode motion as tokens on a uniform temporal grid, where every token spans the same fixed time window. Human motion, however, is temporally heterogeneous: slowly evolving global trajectories coexist with rapid transient events such as foot contacts and joint impulses. Forcing such multi-scale dynamics onto tokens of identical temporal resolution entangles motion frequencies, leaving slow regions redundant while smoothing out the rapid details that distinguish realistic motion. We propose \textbf{FreqMo}, a scale-adaptive motion representation that decomposes motion into wavelet frequency bands, separating dynamics across temporal scales while preserving temporal localization and exact reconstruction. Unified Frequency Residual Quantization (UFRQ) then encodes all bands within a single shared codebook, compressing the token sequence threefold and enabling stable single-stage generation. Experiments show FreqMo attains SOTA fidelity with substantially improved high-frequency preservation, and the same decomposition transfers to continuous diffusion backbones.