Motion Classification Research Articles

Evaluation of the ActiMotus Software to Accurately Classify Postures and Movements in Children Aged 3–14

Background: ActiMotus, a thigh-accelerometer-based software used for the classification of postures and movements (PaMs), has shown high accuracy among adults and school-aged children; however, its accuracy among younger children and potential differences between sexes are unknown. This study aimed to evaluate the accuracy of ActiMotus to measure PaMs among children between 3 and 14 years and to assess if this was influenced by the sex or age of children. Method: Forty-eight children attended a structured ~1-hour data collection session at a laboratory. Thigh acceleration was measured using a SENS accelerometer, which was classified into nine PaMs using the ActiMotus software. Human-coded video recordings of the session provided the ground truth. Results: Based on both F1 scores and balanced accuracy, the highest levels of accuracy were found for lying, sitting, and standing (63.2–88.2%). For walking and running, accuracy measures ranged from 48.0 to 85.8%. The lowest accuracy was observed for classifying stair climbing. We found a higher accuracy for stair climbing among girls compared to boys and for older compared to younger age groups for walking, running, and stair climbing. Conclusions: ActiMotus could accurately detect lying, sitting, and standing among children. The software could be improved for classifying walking, running, and stair climbing, particularly among younger children.

Automated Unmanned Aerial System for Camera-Based Semi-Automatic Triage Categorization in Mass Casualty Incidents

Using drones to obtain vital signs during mass-casualty incidents can be extremely helpful for first responders. Thanks to technological advancements, vital parameters can now be remotely assessed rapidly and robustly. This motivates the development of an automated unmanned aerial system (UAS) for patient triage, combining methods for the automated detection of respiratory-related movements and automatic classification of body movements and body poses with an already published algorithm for drone-based heart rate estimation. A novel UAS-based triage algorithm using UAS-assessed vital parameters is proposed alongside a robust UAS-based respiratory rate assessment and pose classification algorithm. A pilot concept study involving 15 subjects and 30 vital sign measurements under outdoor conditions shows that with our approach, an overall triage classification accuracy of 89% and an F1 score of 0.94 can be achieved, demonstrating its basic feasibility.

Clinical Utility of Neurophysiologic Classification (and Declassification) of Myoclonus.

Movement clinical neurophysiology studies can distinguish myoclonus, tremor, and other jerky movements; however, there has been limited demonstration of their real-world clinical impact. The aim was to investigate movement study utility in clarifying movement classification and guiding patient management. A retrospective study of myoclonus-related movement studies was performed. Of 262 patients referred for consideration of myoclonus, 105 (40%) had myoclonus, 156 (59%) had no myoclonus (the commonest alternative classifications were functional jerks and tremor), and 1 was uncertain. An additional 29 studies identified myoclonus without prior clinical suspicion. A total of 119 of 134 (89%) myoclonus patients had a specific neurophysiologic subtype identified, most commonly cortical (64, 54%). Diagnostic differential narrowed in 60% of patients, and a new diagnosis was made in 42 (14%) patients. Medication changes were made in 151 patients (52%), with improvement in 35 of 51 (67%) with follow-up. Movement studies effectively determined movement classification and identified unsuspected myoclonus, leading to changes in diagnosis and management. © 2024 The Author(s). Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Machine Learning for Upper Limb Flexion Movement Classification

Introduction: This paper introduces the development and validation of a Machine Learning (ML) program aimed at discerning smooth and jittery arm movements during flexion/extension exercises. Objective: The study compares the efficacy of three classification algorithms—K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Logistic Regression—in differentiating flexion/extension movements with and without added weight. Method: Using a quasi-experimental design, participants voluntarily performed the exercise under two conditions: with and without a 5-kilogram dumbbell. Meticulous frame-by-frame extraction of movement parameters informed the data collection process. Results: Biomechanical analysis identified key features (minAngle, coefTrajectory, maxJerk, avgAcceleration, and frames) relevant for algorithm training. Post-normalization, KNN, Logistic Regression, and SVM demonstrated robust validation performance through metrics and confusion matrices. Detection of a userdependent data leak prompted a user-specific validation approach. Conclusion: This research amalgamates biomechanics and ML, yielding insights into algorithmic performance for detecting weighted exercises. Robust validation is crucial for ensuring the generalizability of classification models in real-world scenarios.

Fine-grained Fidgety Movement Classification using Active Learning.

Typically developing infants, between the corrected age of 9-20 weeks, produce fidgety movements. These movements can be identified with the General Movement Assessment, but their identification requires trained professionals to conduct the assessment from video recordings. Since trained professionals are expensive and their demand may be higher than their availability, computer vision-based solutions have been developed to assist practitioners. However, most solutions to date treat the problem as a direct mapping from video to infant status, without modeling fidgety movements throughout the video. To address that, we propose to directly model infants' short movements and classify them as fidgety or non-fidgety. In this way, we model the explanatory factor behind the infant's status and improve model interpretability. The issue with our proposal is that labels for an infant's short movements are not available, which precludes us to train such a model. We overcome this issue with active learning. Active learning is a framework that minimizes the amount of labeled data required to train a model, by only labeling examples that are considered "informative" to the model. The assumption is that a model trained on informative examples reaches a higher performance level than a model trained with randomly selected examples. We validate our framework by modeling the movements of infants' hips on two representative cohorts: typically developing and at-risk infants. Our results show that active learning is suitable to our problem and that it works adequately even when the models are trained with labels provided by a novice annotator.

Attitude motion classification of resident space objects using light curve spectral analysis

The characterization of Earth-orbiting objects in terms of attitude motion, material composition, and shape is crucial for Space Domain Awareness. In this respect, light curves, i.e., graphs showing the brightness variation of space objects over a specific time interval, can be used to infer both their dynamical and physical properties. However, such light curve inversion problem is challenging due to measurement noise, data gaps, and the intrinsic difficulty in separating the multiple effects impacting on the object’s brightness. Existing light curve inversion methods rely on estimation filters exploiting data fusion, Machine Learning approaches, or are based on the analysis of the light-curve frequency spectra. Many of these approaches make assumptions about a-priori knowledge of target object’s characteristics, such as shape, size, surface reflective properties, and/or observation parameters, e.g., target’s orbit and observation geometry. In this framework, this paper presents an innovative architecture for the classification of the attitude motion of space objects using only photometric light curve data, not requiring any a-priori assumption or knowledge. The proposed architecture exploits the Lomb-Scargle Periodogram algorithm to retrieve a frequency spectrum from light curve measurements, and then performs the classification by analyzing the spectrum characteristics testing different candidate rotation periods through the Phase Dispersion Minimization method. The proposed classification algorithm is first tested using a light curve simulator, in which any complex geometric model and different surface properties can be generated. Finally, the method is applied to real light curves retrieved from a publicly available database to assess the classification performance.

A Novel Aerial-Aquatic Unmanned Vehicle Using Flapping Wings for Underwater Propulsion.

Aerial-aquatic unmanned vehicles are a combination of unmanned aerial vehicles and unmanned submersibles, capable of conducting patrols in both the air and underwater domains. This article introduces a novel aerial-aquatic unmanned vehicle that integrates fixed-wing configuration and flapping-wing configuration. In order to improve the low efficiency of the classic diagonal motion trajectory, this paper proposed an improved diagonal motion trajectory based on joint optimization of the stroke angle and angle of attack curve. The proposed method has been verified through simulations and experiments. A prototype was developed and experiments were completed, both indoors and outdoors, wherein the system's transmedium transition capability and flapping propulsion performance were comprehensively validated. Additionally, utilizing flapping propulsion, an average underwater propulsion speed of 0.92 m/s was achieved.

Physical demands of collegiate basketball practice: a preliminary report on novel methods and metrics.

Knowing the specific physical demands of basketball players can provide useful information for clinical decision making when rehabilitating athletes following injury. The purpose of this observational study was to describe the physical demands of basketball play at the Division I collegiate level using video-based time-motion analysis and introduce a time-efficient alternative method of quantifying demands. Eleven NCAA Division I basketball players (6M, 5F; 4 guards, 4 centers, 3 forwards) participated in the study. Video footage was collected from four practices (2 men's, 2 women's) and used to quantify the types and frequencies of player movements based on definitions from seminal work. A second and simpler method was also used to classify movement. A two-way ANOVA was used to assess significant differences in movement by team (men's, women's) and position (guard, forward, center). There were significant differences in counts of stand/walk (p < 0.001), jog (p = 0.012), run (p = 0.001), stride/sprint (p = 0.04), and medium-intensity shuffling (p < 0.001) per minute and proportion of practice time spent in bodyweight (p < 0.001) or above-bodyweight (p < 0.001) loading between teams. There were significant differences for jog (p = 0.001) and transition (p = 0.07) rates across positions. Position and team are important considerations for rehabilitation and return-to-sport clearance. Quantification of these demands can be reliably acquired through video analysis using a simplified method (estimated foot load) or using traditional methods of movement classification and counts, particularly when applying descriptors that better capture the current style of play.

Classification of movements of collegiate female soccer players using inertial measurement units

This study aimed to classify the movements of female soccer players during matches using raw data measured by inertial measurement units (IMUs). Twelve collegiate female soccer players were equipped with IMUs (100 Hz), and raw triaxial acceleration data from eight official matches were analyzed. The measurement data were separated every 3 s, and a Fast Fourier Transform (FFT) was performed. After FFT, the Euclidean distances between the data when the players were in the stationary state and other states were calculated to classify the movements of the players using the k-means method. The data of the clustering numbers classified as the stationary state were eliminated after analyzing the movements of players by video filming the matches. After classification, the average Euclidean distances between the stationary state and other movements were calculated. Consequently, the results showed that the upward and downward directions of the raw data affected the classification. Using the methods of this study, it was also shown that the distribution of Euclidean distances differed from player to player. Our findings indicate that the method used in this study can be used to classify and characterize the movements of female collegiate soccer players.

Detection of Movement-Related Brain Activity Associated with Hand and Tongue Movements from Single-Trial Around-Ear EEG.

Movement intentions of motor impaired individuals can be detected in laboratory settings via electroencephalography Brain-Computer Interfaces (EEG-BCIs) and used for motor rehabilitation and external system control. The real-world BCI use is limited by the costly, time-consuming, obtrusive, and uncomfortable setup of scalp EEG. Ear-EEG offers a faster, more convenient, and more aesthetic setup for recording EEG, but previous work using expensive amplifiers detected motor intentions at chance level. This study investigates the feasibility of a low-cost ear-EEG BCI for the detection of tongue and hand movements for rehabilitation and control purposes. In this study, ten able-bodied participants performed 100 right wrist extensions and 100 tongue-palate movements while three channels of EEG were recorded around the left ear. Offline movement vs. idle activity classification of ear-EEG was performed using temporal and spectral features classified with Random Forest, Support Vector Machine, K-Nearest Neighbours, and Linear Discriminant Analysis in three scenarios: Hand (rehabilitation purpose), hand (control purpose), and tongue (control purpose). The classification accuracies reached 70%, 73%, and 83%, respectively, which was significantly higher than chance level. These results suggest that a low-cost ear-EEG BCI can detect movement intentions for rehabilitation and control purposes. Future studies should include online BCI use with the intended user group in real-life settings.

To Manipulate or Not? Management of Pediatric Knee Arthrofibrosis Following Operative Fixation of Tibial Spine Fractures

ObjectivesKnee arthrofibrosis is a common complication after surgical fixation of tibial spine fractures. However, there is no standardized treatment modality for resultant arthrofibrosis, with some surgeons electing for non-operative management, while others prefer manipulation under anesthesia with a possible arthroscopic lysis of adhesions, if indicated. To better understand indications and outcomes from these treatment modalities for arthrofibrosis, we examined patients treated by both approaches. MethodsWe performed a multicenter retrospective case series of patients with arthrofibrosis after tibial spine fracture surgery. Arthrofibrosis was defined as loss of knee extension ≥ 10.0° and/or knee flexion ≥ 25.0° compared to the contralateral, uninjured knee at three months following index surgery. Patients were organized into two cohorts: non-operative and operative treatment. Data was collected for demographics, fracture classification, concomitant injuries, treatment timeline, and knee range of motion. ResultsThere were 16 patients in the operative group and 10 patients in the non-operative group. At the time of diagnosis, the operative group had a mean 10° larger flexion deficit compared to the non-operative group. Both treatment modalities resulted in similar ranges of motion at terminal evaluation. Final flexion was recorded as 130° in the operative group and 127° in the non-operative group. Final extension deficits were 3° in both cohorts. ConclusionsOperative and non-operative treatment modalities can be effective in management of knee arthrofibrosis after fixation of tibial spine fractures. Non-operative treatment may be more suitable for milder range of motion deficits, but further research is necessary to guide clinical practice. Level of EvidenceLevel IV, Case Series;

Movements classification system for transhumeral amputees using myoelectric signals

The extent of loss of functionality in people with transhumeral amputation or elbow disarticulation is considered highly compromised. The person with this impairment loses the primary muscle involved in the movement of the hand and fingers, and the elbow flexion/extension. In this sense, the development of myoelectric prostheses must include mechanisms that replicate these movements. Moreover, the control system for controlling a prosthesis must be robust and intuitive, where pattern recognition and artificial intelligence techniques represent an important and increasingly explored alternative. This study aims to develop a movement intention classification system based on myoelectric signals (MES) obtained from an elbow disarticulation amputee. Different features were extracted from the MES as the Mean absolute value (MAV), zero crossings (ZC), slope sign changes (SSC), waveform length (WL), and autoregressive (AR) coefficients. Machine learning classifiers (Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Random Forests (RF), and Multilayer Perceptron (MLP)) were used to predict seven gestures: elbow flexion (EF), elbow extension (EE), forearm pronation (FP), forearm supination (FS), hand opening (HO), hand closing (HC), and rest (R). The results obtained with LDA, QDA, KNN, MLP, SVM, and RF were 60.84%, 71.05%, 77.10%, 78.06%, 78.57%, and 79.72% of accuracy, respectively. These results are promising and will allow the development of a myoelectric prosthesis control system for people with transhumeral amputation or elbow disarticulation.

Abnormal activation patterns in MT+ during visual motion perception in major depressive disorder.

Previous studies have found that patients with Major Depressive Disorder (MDD) exhibit impaired visual motion perception capabilities, and multi-level abnormalities in the human middle temporal complex (MT+), a key brain area for processing visual motion information. However, the brain activity pattern of MDD patients during the perception of visual motion information is currently unclear. In order to study the effect of depression on the activity and functional connectivity (FC) of MT+ during the perception of visual motion information, we conducted a study combining task-state fMRI and psychophysical paradigm to compare MDD patients and healthy control (HC). Duration threshold was examined through a visual motion perception psychophysical experiment. In addition, a classic block-design grating motion task was utilized for fMRI scanning of 24 MDD patients and 25 HC. The grating moved randomly in one of eight directions. We examined the neural activation under visual stimulation conditions compared to the baseline and FC. Compared to HC group, MDD patients exhibited increased duration threshold. During the task, MDD patients showed decreased beta value and percent signal change in left and right MT+. In the sample comprising MDD and HC, there was a significant negative correlation between beta value in right MT+ and duration threshold. And in MDD group, activation in MT+ were significantly correlated with retardation score. Notably, no such differences in activation were observed in primary visual cortex (V1). Furthermore, when left MT+ served as the seed region, compared to the HC, MDD group showed increased FC with right calcarine fissure and surrounding cortex and decreased FC with left precuneus. Overall, the findings of this study highlight that the visual motion perception function impairment in MDD patients relates to abnormal activation patterns in MT+, and task-related activity are significantly connected to the retardation symptoms of the disease. This not only provides insights into the potential neurobiological mechanisms behind visual motion perception disorder in MDD patients from the aspect of task-related brain activity, but also supports the importance of MT+ as a candidate biomarker region for MDD.

Independent Vector Analysis for Feature Extraction in Motor Imagery Classification.

Independent vector analysis (IVA) can be viewed as an extension of independent component analysis (ICA) to multiple datasets. It exploits the statistical dependency between different datasets through mutual information. In the context of motor imagery classification based on electroencephalogram (EEG) signals for the brain-computer interface (BCI), several methods have been proposed to extract features efficiently, mainly based on common spatial patterns, filter banks, and deep learning. However, most methods use only one dataset at a time, which may not be sufficient for dealing with a multi-source retrieving problem in certain scenarios. From this perspective, this paper proposes an original approach for feature extraction through multiple datasets based on IVA to improve the classification of EEG-based motor imagery movements. The IVA components were used as features to classify imagined movements using consolidated classifiers (support vector machines and K-nearest neighbors) and deep classifiers (EEGNet and EEGInception). The results show an interesting performance concerning the clustering of MI-based BCI patients, and the proposed method reached an average accuracy of 86.7%.

Correction to: Pose pattern mining using transformer for motion classification

Correction to: Pose pattern mining using transformer for motion classification

Transient motion classification and segment analysis of diffusive trajectories of G proteins and coupled-receptors in a living cell

The molecular movement in single particle tracking (SPT) experiments shows a crucial role of diffusion in many biological processes such as signaling, cellular organization, transport mechanisms, and more. The SPT analysis detects not only classical Brownian motion but diffusion with other features. These include directed diffusion and confined motion. The behavior remains a challenging problem for several reasons. Due to the action of many physical processes, random trajectories of cellular molecules are segmented in different diffusive modes. Often their study requires sophisticated algorithms for the analysis of statistical properties. In this paper we consider the segment analysis for trajectories of G proteins and coupled-receptors in living cells. Their movement is often transient and switches among free diffusion, confined diffusion, directed diffusion, and immobility. Moreover, the confined segments can have both Gaussian and non-Gaussian statistics. The types of alternation of diffusive modes along the trajectories of G proteins and coupled-receptors are analyzed.

PRF: A Program Reuse Framework for Automated Programming by Learning from Existing Robot Programs

This paper explores the problem of automated robot program generation from limited historical data when neither accurate geometric environmental models nor online vision feedback are available. The Program Reuse Framework (PRF) is developed, which uses expert-defined motion classes, a novel data structure introduced in this work, to learn affordances, workspaces, and skills from historical data. Historical data comprise raw robot joint trajectories and descriptions of the robot task being completed. Given new tasks, motion classes are then used again to formulate an optimization problem capable of generating new open-loop, skill-based programs to complete the tasks. To cope with a lack of geometric models, a technique to learn safe workspaces from demonstrations is developed, allowing the risk of new programs to be estimated before execution. A new learnable motion primitive for redundant manipulators is introduced, called a redundancy dynamical movement primitive, which enables new end-effector goals to be reached while mimicking the whole-arm behavior of a demonstration. A mobile manipulator part transportation task is used throughout to illustrate each step of the framework.

Enhanced Classification of Human Fall and Sit Motions Using Ultra-Wideband Radar and Hidden Markov Models

In this study, we address the challenge of accurately classifying human movements in complex environments using sensor data. We analyze both video and radar data to tackle this problem. From video sequences, we extract temporal characteristics using techniques such as motion history images (MHI) and Hu moments, which capture the dynamic aspects of movement. Radar data are processed through principal component analysis (PCA) to identify unique detection signatures. We refine these features using k-means clustering and employ them to train hidden Markov models (HMMs). These models are tailored to distinguish between distinct movements, specifically focusing on differentiating sitting from falling motions. Our experimental findings reveal that integrating video-derived and radar-derived features significantly improves the accuracy of motion classification. Specifically, the combined approach enhanced the precision of detecting sitting motions by over 10% compared to using single-modality data. This integrated method not only boosts classification accuracy but also extends the practical applicability of motion detection systems in diverse real-world scenarios, such as healthcare monitoring and emergency response systems.

Human activity recognition system using wearable accelerometers for classification of leg movements: a first, detailed approach

Abstract. A human activity recognition (HAR) system carried by masseurs for controlling a therapy table via different movements of legs or hip is studied. This work starts with a survey on HAR systems using the sensor position named “trouser pockets”. Afterwards, in the experiments, the impacts of different hardware systems, numbers of subjects, data generation processes (online streams/offline data snippets), sensor positions, sampling rates, sliding window sizes and shifts, feature sets, feature elimination processes, operating legs, tag orientations, classification processes (concerning method, parameters and an additional smoothing process), numbers of activities, training databases, and the use of a preceding teaching process on the classification accuracy are examined to get a thorough understanding of the variables influencing the classification quality. Besides the impacts of different adjustable parameters, this study also serves as an advisor for the implementation of classification tasks. The proposed system has three operating classes: do nothing, pump therapy table up or pump therapy table down. The first operating class consists of three activity classes (go, run, massage) such that the whole classification process exists with five classes. Finally, using online data streams, a classification accuracy of 98 % could be achieved for one skilled subject and about 90 % for one randomly chosen subject (mean of 1 skilled and 11 unskilled subjects). With the LOSO (leave-one-subject-out) technique for 12 subjects, up to 86 % can be attained. With our offline data approach, we get accuracies of 98 % for 12 subjects and up to 100 % for 1 skilled subject.

Enhancing Radar Image Classification with Autoencoder-CNN Hybrid System

The tracking, analysis, and classification of human movements can be crucial, particularly in areas such as elderly care, healthcare, and infant care. Typically, such tracking is done remotely with cameras. However, radar systems have emerged as significant methods and tools for these tasks due to their advantages such as privacy, wireless operation, and the ability to work through walls. By converting reflected radar signals from targets into images, human activities can be classified using powerful classification tools like deep learning. In this study, range-Doppler images of behind-the-wall human movements obtained with a radar system consisting of one transmitter and four receiver antennas were classified. Since the data collected from the four receiver antennas are in different positions, the collected reflection signals also differ. The signals collected with the range-time matrix content were divided into positive and negative parts, resulting in eight images from the four antennas. Instead of using all the data in CNN training, the images were first subjected to a reconstruction process with an autoencoder to reduce differences. As a result, it was observed that reconstructing the images with an autoencoder before classification with CNN increased the classification success. In conclusion, it was observed that the classification success of radar images can be increased by using a hybrid system with an autoencoder to reconstruct the images before classification with CNN.