Nonlinear dance motion analysis and motion editing using Hilbert-Huang transform

TL;DR

This paper introduces a novel method using the Hilbert-Huang transform to decompose, analyze, and edit complex dance motions, enabling the creation of new dance sequences through motion blending and modification.

Contribution

It presents a new framework leveraging HHT for nonlinear motion analysis and editing, specifically applied to dance motions, which is more effective than traditional Fourier or wavelet methods.

Findings

Successfully decomposed dance motions into IMFs

Enabled blending and modification of dance primitives

Produced new dance motions through motion editing

Abstract

Human motions (especially dance motions) are very noisy, and it is hard to analyze and edit the motions. To resolve this problem, we propose a new method to decompose and modify the motions using the Hilbert-Huang transform (HHT). First, HHT decomposes a chromatic signal into "monochromatic" signals that are the so-called Intrinsic Mode Functions (IMFs) using an Empirical Mode Decomposition (EMD) [6]. After applying the Hilbert Transform to each IMF, the instantaneous frequencies of the "monochromatic" signals can be obtained. The HHT has the advantage to analyze non-stationary and nonlinear signals such as human-joint-motions over FFT or Wavelet transform. In the present paper, we propose a new framework to analyze and extract some new features from a famous Japanese threesome pop singer group called "Perfume", and compare it with Waltz and Salsa dance. Using the EMD, their dance motions can be decomposed into motion (choreographic) primitives or IMFs. Therefore we can scale, combine, subtract, exchange, and modify those IMFs, and can blend them into new dance motions self-consistently. Our analysis and framework can lead to a motion editing and blending method to create a new dance motion from different dance motions.