Detecting the relevance to performance of whole-body movements

Abstract

Goal-directed whole-body movements are fundamental in our daily life, sports, music, art, and other activities. Goal-directed movements have been intensively investigated by focusing on simplified movements (e.g., arm-reaching movements or eye movements); however, the nature of goal-directed whole-body movements has not been sufficiently investigated because of the high-dimensional nonlinear dynamics and redundancy inherent in whole-body motion. One open question is how to overcome high-dimensional nonlinear dynamics and redundancy to achieve the desired performance. It is possible to approach the question by quantifying how the motions of each body part at each time point contribute to movement performance. Nevertheless, it is difficult to identify an explicit relation between each motion element (the motion of each body part at each time point) and performance as a result of the high-dimensional nonlinear dynamics and redundancy inherent in whole-body motion. The current study proposes a data-driven approach to quantify the relevance of each motion element to the performance. The current findings indicate that linear regression may be used to quantify this relevance without considering the high-dimensional nonlinear dynamics of whole-body motion.

Detection of task-relevant and task-irrelevant motion sequences: application to motor adaptation in goal-directed and whole-body movements

“…The current study relied on linear regression to determine the relationship between motion data X ∈ R T×D (the current study focused on the temporal sequence of joint angles and angular velocities) and performance data d ∈ R T ×1 following h = Xw , where T and D denoted the number of trials and the number variables in the motion data, h ∈ R T ×1 is the predicted performance, and w ∈ R D× 1 is the best linear coefficients to predict the performance [30]. X t , the t th row of X or the motion data at the t th trial, consists of vectorized motion data (e.g., after measuring joint angles of knee q k ,t ∈ R 1 ×F and hip q h ,t ∈ R 1 ×F for F time frames at the t th trial, X t = ( q k ,t , q h ,t )).…”

“…We needed to validate the ridge regression in the current experimental setting before evaluating variability. Notably, we have already validated the efficiency of ridge regression to predict performance not only in jumping movements but also in throwing movements [30].…”

“…Our framework is likely sensitive to how to choose the coordinate; however, in contrast to the UCM framework, our method enables the selection of the appropriate coordinate for discussing the relationship between motion and performance based on predictive power. Although we considered one-dimensional performance in the current study (i.e., jumping height), two-dimensional performance requires the definition of an appropriate coordinate to discuss performance [30, 36]. Predictive power plays a vital role in selecting the proper coordinate not only in motion but also in the performance space [30].…”

“…Here, we propose a flexible and straightforward machine learning technique that can evaluate movement variability by unifying the advantages of previous techniques: our framework can evaluate not only task-relevant and task-irrelevant variabilities even when averaged kinematics or task parameters change (e.g., before, during, and after motor learning) while considering motion sequence and task function but also how each synergy is relevant to task performance by extending PCA. The current study relied on a ridge regression [29], a linear regression technique that is robust in the presence of measurement noise, has a definite relation to PCA, and can evaluate how the motion of each body part at each time is relevant to task performance in a data-driven manner [30]. Our technique can thus enable the identification of task functions in a data-driven manner without any explicit function such as the parabola or the forward kinematics.…”

Detecting task-dependent modulation of spatiotemporal module via tensor decomposition: application to kinematics and EMG data for walking and running at various speed

“…For extracting such error-correction-related modules, the tensor decomposition can be effectively used after being combined with regression frameworks [38]. A regression framework enables us to extract the hidden information relevant to target values [28, 39]. Because several studies have focused on adaptation [35–37, 40–43] and spatiotemporal modules [2–6, 11] separately, the relation between those concepts has only been investigated in a few studies [10, 16, 28].…”