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

Abstract

Detecting human motion and predicting human intentions by analyzing body signals are challenging but fundamental steps for the implementation of applications presenting human–robot interaction in different contexts, such as robotic rehabilitation in clinical environments, or collaborative robots in industrial fields. Machine learning techniques (MLT) can face the limit of small data amounts, typical of this kind of applications. This paper studies the illustrative case of the reaching movement in 10 healthy subjects and 21 post-stroke patients, comparing the performance of linear discriminant analysis (LDA) and random forest (RF) in: (i) predicting the subject’s intention of moving towards a specific direction among a set of possible choices, (ii) detecting if the subject is moving according to a healthy or pathological pattern, and in the case of discriminating the damage location (left or right hemisphere). Data were captured with wearable electromagnetic sensors, and a sub-section of the acquired signals was required for the analyses. The possibility of detecting with which arm (left or right hand) the motion was performed, and the sensitivity of the MLT to variations in the length of the signal sub-section were also evaluated. LDA and RF prediction accuracies were compared: Accuracy improves when only healthy subjects or longer signals portions are considered up to 11% and at least 10%, respectively. RF reveals better estimation performance both as intention predictor (on average 59.91% versus the 62.19% of LDA), and health condition detector (over 90% in all the tests).

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

“…Compared to previous research in this field, it has been demonstrated that linear discriminant analysis (LDA) and random forest (RF) can effectively predict a subject's intention to initiate movement. Notably, the average prediction transition time was approximately 31 × 10 −4 seconds and 11 × 10 −5 seconds, respectively [31,33]. It is important to note that these studies predominantly focused on the upper limbs and arms only, characterized by relatively simple joint freedom and kinematic parameters.…”

“…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].…”

A Low-Cost Inertial Measurement Unit Motion Capture System for Operation Posture Collection and Recognition

“…Machine learning is commonly used in human motion recognition research, for example, the support vector machines classification model [ 16 ], Markov model [ 17 ], and random forest (RF) [ 18 ]. In the past few years, deep learning algorithms have found extensive applications in the realm of human motion recognition [ 19 ], demonstrating superior recognition performance compared to traditional algorithms [ 20 , 21 ].…”

“…Moreover, for the collection of human body movements, the sampling rate and accuracy within our solution fall within acceptable ranges, and the interface and the battery life employed in our solution are also sufficient for human motion capture. Machine learning is commonly used in human motion recognition research, for example, the support vector machines classification model [16], Markov model [17], and random forest (RF) [18]. In the past few years, deep learning algorithms have found extensive applications in the realm of human motion recognition [19], demonstrating superior recognition performance compared to traditional algorithms [20,21].…”

Machine Learning for Human Motion Intention Detection

“…The CNN is still a static-data-based network and the RNN result is still geared toward large motion trajectories. The result by Ragni et al [ 16 ] addressed intent recognition, but was limited to predicting the subject’s choice in three possible ways. Additionally, much of their work was dedicated to deciding if the subject was healthy or a post-stroke patient.…”