TY - JOUR

T1 - Global Position Prediction for Interactive Motion Capture

AU - Schreiner, Paul

AU - Perepichka, Maksym

AU - Lewis, Hayden

AU - Darkner, Sune

AU - Kry, Paul G.

AU - Erleben, Kenny

AU - Zordan, Victor B.

PY - 2021

Y1 - 2021

N2 - We present a method for reconstructing the global position of motion capture where position sensing is poor or unavailable. Capture systems, such as IMU suits, can provide excellent pose and orientation data of a capture subject, but otherwise need post processing to estimate global position. We propose a solution that trains a neural network to predict, in real-time, the height and body displacement given a short window of pose and orientation data. Our training dataset contains pre-recorded data with global positions from many different capture subjects, performing a wide variety of activities in order to broadly train a network to estimate on like and unseen activities. We compare training on two network architectures, a universal network (u-net) and a traditional convolutional neural network (CNN) - observing better error properties for the u-net in our results. We also evaluate our method for different classes of motion. We observe high quality results for motion examples with good representation in specialized datasets, while general performance appears better in a more broadly sampled dataset when input motions are far from training examples.

AB - We present a method for reconstructing the global position of motion capture where position sensing is poor or unavailable. Capture systems, such as IMU suits, can provide excellent pose and orientation data of a capture subject, but otherwise need post processing to estimate global position. We propose a solution that trains a neural network to predict, in real-time, the height and body displacement given a short window of pose and orientation data. Our training dataset contains pre-recorded data with global positions from many different capture subjects, performing a wide variety of activities in order to broadly train a network to estimate on like and unseen activities. We compare training on two network architectures, a universal network (u-net) and a traditional convolutional neural network (CNN) - observing better error properties for the u-net in our results. We also evaluate our method for different classes of motion. We observe high quality results for motion examples with good representation in specialized datasets, while general performance appears better in a more broadly sampled dataset when input motions are far from training examples.

KW - IMU

KW - motion capture

KW - neural networks

UR - http://www.scopus.com/inward/record.url?scp=85116454715&partnerID=8YFLogxK

U2 - 10.1145/3479985

DO - 10.1145/3479985

M3 - Journal article

AN - SCOPUS:85116454715

VL - 4

SP - 1

EP - 16

JO - Proceedings of the ACM on Computer Graphics and Interactive Techniques

JF - Proceedings of the ACM on Computer Graphics and Interactive Techniques

SN - 2577-6193

IS - 3

M1 - 3479985

ER -