Inclusive Human Intention Prediction with Wearable Sensors: Machine Learning Techniques for the Reaching Task Use Case

Abstract

Human intentions prediction is gaining importance with the increase in human–robot interaction challenges in several contexts, such as industrial and clinical. This paper compares Linear Discriminant Analysis (LDA) and Random Forest (RF) performance in predicting the intention of moving towards a target during reaching movements on ten subjects wearing four electromagnetic sensors. LDA and RF prediction accuracy is compared to observation-sample dimension and noise presence, training and prediction time. Both algorithms achieved good accuracy, which improves as the sample dimension increases, although LDA presents better results for the current dataset.

Intention Prediction and Human Health Condition Detection in Reaching Tasks with Machine Learning Techniques

“…During each session, at first a preliminary trial was carried out to familiarize the subject with the procedure, then the subject was asked to perform six repetitions of unilateral sitting reaching movement: three cycles with the left arm, and three with the right arm. As described in Robertson et al [19] and Archetti et al [30], the initial condition consisted in the subject with the hand placed on a red cross placed on the table plane in line with the shoulder, the forearm in mid-prone, the elbow flexed to 90 • , and the humerus positioned along the vertical direction. In each repetition, the subject was asked to touch the target identified by the operator, among a set of possible pre-defined positions, which depicts combinations of the three directions left, center, and right, of the two quotes high and low, and of the two distances proximal and distal.…”

“…To detect starting and ending points of the reaching movement, the absolute value of the velocity of the hand was analyzed. The absolute value of the hand position was computed as the vectorial composition of the signal components along the three directions X, Y, and Z, and the hand velocity was numerically evaluated according to a custom two-point derivative approximation [30]. This signal was then filtered to remove noise with a fourth-order zerophase low-pass Butterworth filter, according to literature indications [31,32].…”

“…The performance of LDA and RF are compared for the specific analysis of the reaching movement; (iii). With particular reference to the previous work of Archetti et al [30], the analysis of a dataset which also includes data collected from post-stroke patients on one side allowed a more robust evaluation of the performance of LDA and RF as intention predictors, and on the other side enabled the implementation of a new level of analysis, evaluating the performance of LDA and RF as health condition detectors.…”

Human Posture Transition-Time Detection Based upon Inertial Measurement Unit and Long Short-Term Memory Neural Networks

“…Additionally, there has also been much research using neural network (NN) techniques such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), and deep learning [15,[29][30][31]. Archetti et al and Ragni et al compared performances between linear discriminant analysis (LDA) and random forest (RF) in predicting the intended reaching of the target with subjects wearing electromagnetic sensors [32,33]. Li et al used action recognition, action prediction, and posture-change detection to predict the pitcher's choice of one of the nine-square divisions by capturing and analyzing the pitcher's RGB image and optical flow [34].…”