Human Motion Classification Research Articles

Wavelet Based Denoising of the Simulated Chest Wall Motion Detected by SFCW Radar

Low power and compact radars have emerged with the development of electronic technology. This has enabled the use of radars in indoor environments and the realization of many applications. The detection, tracking and classification of human movements by radar are among the remarkable applications. Contactless detection of human vital signs improves the quality of life of patients being kept under observation and facilitates the work of experts. In this study, it was simulated that the movement of the chest wall was modeled and detected by the SFCW radar. Gaussian, Rician and uniformly distributed random noise types were added to the modeled chest motion at different levels. The noisy signal obtained at the receiver is denoised with different mother wavelet functions and the performances of these functions are presented comparatively.

A Wrist Worn Acceleration Based Human Motion Analysis and Classification for Ambient Smart Home System

In recent years, health-care industry has received a major boost due to sensors i.e., accelerometers, magnetometers etc., which allow its user to get instant updates about their current health status in indoor/outdoor environments. The real driving force behind the usage of accelerometer has been the fitness industry but it also holds a prominent place in ambient smart home to monitor resident’s life-style. In this paper, we proposed a novel triaxial accelerometer-based human motion detection and recognition system using multiple features and random forest. Triaxial signals have been statistically processed to produce worthy features like variance, positive and negative peaks, and signal magnitude features. The proposed model was evaluated over HMP recognition data sets and achieved satisfactory recognition accuracy of 85.17%. The proposed system is directly applicable to any elderly/children health monitoring system, 3D animated games/movies and examining the indoor behaviors of people at home, malls and offices.

Unsupervised Domain Adaptation for Micro-Doppler Human Motion Classification via Feature Fusion

Micro-Doppler-based human motion classification has become a topical area of research recently. However, the current research is limited by the lack of labeled training data. Domain adaptation, namely, the ability to take advantage of knowledge from an available source data set and apply it to an unlabeled target data set, is useful in this situation. A typical strategy for this transfer learning technique is to extract domain-invariant feature representations. In this letter, an unsupervised domain adaptation method for micro-Doppler classification is proposed. Given no available measurement training samples, we creatively utilize the motion capture database as an auxiliary and adapt its interior knowledge to the measurement data set. To achieve domain-invariant features, three types of features are extracted and fused including low-level deep features from the convolutional neural network, empirical features, and statistical features. After feature fusion, a $k$ -nearest neighbor classifier is applied to the measurement data to classify seven human activities. Experimental results show that our approach outperforms several state-of-the-art unsupervised domain adaptation methods. The impact of the output from different convolution layers is further investigated, and ablation studies of the efficacy of each feature are also carried out in this letter.

Research on de-motion blur image processing based on deep learning

In recent years, with the rapid development of computer technology and network technology, computer vision has been widely used in various scientific fields. Human motion recognition, as an important branch of computer vision, is essentially to classify human motion information in motion images correctly. It has great significance in intelligent monitoring and security, human-computer interaction, motion analysis and other fields. At present, there are still some problems in human motion recognition methods. Firstly, how to extract and characterize the motion information in images has been one of the difficulties in this field; secondly, with the appearance of kinect and other depth cameras, researchers have provided the depth information of human motion images, and how to effectively use these depth information to achieve human motion recognition and classification is also an important research issue; finally, when the amount of sample data is small, how to use the deep learning network model to achieve a higher human motion recognition rate? Based on UTD-MHAD database, this paper studies the human motion recognition of RGB image and depth image captured simultaneously by kinect, and carries out relevant discussion and analysis on the above problems, using micro-inertial sensors (MTi-G-700 developed by Xsens and Android mobile phones, tablets and other personal mobile devices come with MEMS gyroscopes and accelerometers) to correct the image to motion blur, build a new mathematical model, use the inertial data obtained by MIMU in a short time to estimate the position, attitude and speed of camera motion, correct the image pixel position, perform image de-motion blur processing, and then perform image processing such as denoising to solve the image motion blur problem. A new algorithm is developed and its science is verified by MATLAB simulation.

Human wrist motion classification using decision tree and principle component analysis

Abstract This study investigates and acts as a trial clinical outcome for human motion and behaviour analysis in consensus of health related quality of life in Malaysia. Both methods were developed in MATLAB platform by managing and analyzing motion data in graphical user interface (GUI) form. By attaching the wearable sensors to each segments of interest, the collected data can be reconstructed in real time basis. Artificial intelligence technique is proposed to classify the subjects’ motions. Data reconstruction technique can be used to regenerate the motion of subjects’ limbs movement in 2D representation. Corresponding tests are performed to validate the accuracy of the application. The application is potentially to be used to access the quality of human motion in hospitals, clinics and human motion research. An experiment was set up in a laboratory environment for data collecting. Subjects were asked to perform three shape drawings using left and right hands. The instruments demonstrate adequate internal consistency of optimum average accuracy of 52 % for decision tree and 80 % for principle component analysis. Results were presented in tabular, image and graphical forms. The expected output of the study can be used to estimate joint angles in predicting Parkinson Disease for early stage symptoms, to track the moving behaviour in clinical gait analysis, to analyse limb movement in detecting resting tremor, to trace wrist motion in regenerating children handwriting difficulty, and to measure joint motion in evaluating patient rehabilitation progress.

Human activity recognition from smart watch sensor data using a hybrid of principal component analysis and random forest algorithm

Background: Detecting of human movements is an important task in various areas such as healthcare, fitness and eldercare. It is now possible to achieve this aim using mobile applications. These applications provide users, doctors and related persons a better understanding about daily physical activities. It can also lead to various useful habits by following the activities of the users in their daily life. In addition, dangerous actions such as the fall of elderly people or young children are identified and necessary precautions are taken as soon as possible. Classification of human motions with motion sensor data is among the current topics of study. Smart watches have these sensors built-in. Thus, it is possible to follow the activities of a user carrying only a smart watch. Methods: The purpose of this work is to detect human movements using smart watch sensor data and machine learning methods. The data are obtained from the accelerometer, gyroscope, step counter and heart rate sensors of the smart watch. The obtained data have been divided into 2 s windows and a data set containing 500 patterns for each class has been created for each class. Results and Discussion: After the features were determined, the data set to which the principal component analysis has been applied was classified by random forest, support vector machine, C4.5 and k-nearest neighbor methods, and their performances were compared. The most successful result was obtained from the random forest method.

Energy Efficiency Optimization Based on Context Awareness in Wearable Computing

The motion capture system of the wireless wearable device can achieve identification and classification of human motion through various classifiers, but wireless data transmission consumes most of the energy of the sensor node. An adaptive wireless transmission control algorithm based on context awareness is proposed realize energy efficiency optimization based on single classifier and multiple classifiers fusion. Firstly, adaptive wireless transmission control method based on fixed single classifier can effectively improve the overall energy efficiency under the limit of classification accuracy. Secondly, adaptive transmission control method of multi-classifier fusion is proposed with adjusting decision threshold based on various classifiers and their characteristics. The experimental results validate our proposed algorithm.

A two-phase human activity classification design using accelerometer data from smartphone

This paper presents a two-phase design for the classification of the human movement activities of running, walking and standing by using accelerometer data logged from smartphone. The first phase of the design is, the two-stage filter applied for the process of raw acceleration data collected from the accelerometer; and the second phase of the design is, the classification of human movements using Logistic Regression with Gradient Descent Algorithm as classifier from the clean data produced by first phase of the design. The raw acceleration data from the tri-axial accelerometer of smartphone contains three axes accelerations. This acceleration data are collected separately for each movement activity and are used for the training and testing of design by using a simulation environment. The activity classification results are promising and show that the proposed design provides an overall accuracy of more than 97% on both train data and the separate independent test data for the classification of human movement activities of running, walking and standing.

A two-phase human activity classification design using accelerometer data from smartphone

This paper presents a two-phase design for the classification of the human movement activities of running, walking and standing by using accelerometer data logged from smartphone. The first phase of the design is, the two-stage filter applied for the process of raw acceleration data collected from the accelerometer; and the second phase of the design is, the classification of human movements using Logistic Regression with Gradient Descent Algorithm as classifier from the clean data produced by first phase of the design. The raw acceleration data from the tri-axial accelerometer of smartphone contains three axes accelerations. This acceleration data are collected separately for each movement activity and are used for the training and testing of design by using a simulation environment. The activity classification results are promising and show that the proposed design provides an overall accuracy of more than 97% on both train data and the separate independent test data for the classification of human movement activities of running, walking and standing.

Suitability of Data Representation Domains in Expressing Human Motion Radar Signals

Human motion recognition plays a key role in various fields including health monitoring. Radar is a type of sensor that has shown remarkable success in the classification of human motions. Different data representation domains have been used for the analysis of radar returns. Each domain provides one aspect of the observed motion not readily discernible in other domains. In this letter, we propose an approach to quantify the domain suitability in representing a human motion. In order to evaluate the proposed approach, we consider the time–frequency domain and the range map. Additionally, based on the demonstrated domain effectiveness, we propose a classification scheme that seeks to incorporate each domain consistent with its offerings. Experimental results show the importance of investigating domain offerings prior to the classification process.

Classification of Multi-class Daily Human Motion using Discriminative Body Parts and Sentence Descriptions

In this paper, we propose a motion model that focuses on the discriminative parts of the human body related to target motions to classify human motions into specific categories, and apply this model to multi-class daily motion classifications. We extend this model to a motion recognition system which generates multiple sentences associated with human motions. The motion model is evaluated with the following four datasets acquired by a Kinect sensor or multiple infrared cameras in a motion capture studio: UCF-kinect; UT-kinect; HDM05-mocap; and YNL-mocap. We also evaluate the sentences generated from the dataset of motion and language pairs. The experimental results indicate that the motion model improves classification accuracy and our approach is better than other state-of-the-art methods for specific datasets, including human–object interactions with variations in the duration of motions, such as daily human motions. We achieve a classification rate of 81.1% for multi-class daily motion classifications in a non cross-subject setting. Additionally, the sentences generated by the motion recognition system are semantically and syntactically appropriate for the description of the target motion, which may lead to human–robot interaction using natural language.

Online human movement classification using wrist-worn wireless sensors

The monitoring and analysis of human motion can provide valuable information for various applications. This work gives a comprehensive overview about existing methods, and a prototype system is also presented, capable of detecting different human arm and body movements using wrist-mounted wireless sensors. The wireless units are equipped with three tri-axial sensors, an accelerometer, a gyroscope, and a magnetometer. Data acquisition was done for multiple activities with the help of the used prototype system. A new online classification algorithm was developed, which enables easy implementation on the used hardware. To explore the optimal configuration, multiple datasets were tested using different feature extraction approaches, sampling frequencies, processing window widths, and used sensor combinations. The applied datasets were constructed using data collected with the help of multiple subjects. Results show that nearly 100% recognition rate can be achieved on training data, while almost 90% can be reached on validation data, which were not utilized during the training of the classifiers. This shows high correlation in the movements of different persons, since the training and validation datasets were constructed of data from different subjects.

Feature Selection Algorithm Considering Trial and Individual Differences for Machine Learning of Human Activity Recognition

In recent years, many studies have been performed on the automatic classification of human body motions based on inertia sensor data using a combination of inertia sensors and machine learning; training data is necessary where sensor data and human body motions correspond to one another. It can be difficult to conduct experiments involving a large number of subjects over an extended time period, because of concern for the fatigue or injury of subjects. Many studies, therefore, allow a small number of subjects to perform repeated body motions subject to classification, to acquire data on which to build training data. Any classifiers constructed using such training data will have some problems associated with generalization errors caused by individual and trial differences. In order to suppress such generalization errors, feature spaces must be obtained that are less likely to generate generalization errors due to individual and trial differences. To obtain such feature spaces, we require indices to evaluate the likelihood of the feature spaces generating generalization errors due to individual and trial errors. This paper, therefore, aims to devise such evaluation indices from the perspectives. The evaluation indices we propose in this paper can be obtained by first constructing acquired data probability distributions that represent individual and trial differences, and then using such probability distributions to calculate any risks of generating generalization errors. We have verified the effectiveness of the proposed evaluation method by applying it to sensor data for butterfly and breaststroke swimming. For the purpose of comparison, we have also applied a few available existing evaluation methods. We have constructed classifiers for butterfly and breaststroke swimming by applying a support vector machine to the feature spaces obtained by the proposed and existing methods. Based on the accuracy verification we conducted with test data, we found that the proposed method produced significantly higher F-measure than the existing methods. This proves that the use of the proposed evaluation indices enables us to obtain a feature space that is less likely to generate generalization errors due to individual and trial differences.

A topological approach for human movement classification and anticipation

The motion capture systems are increasingly used for biomedical purposes. In order to recognize and classify the movements, however whole-body movements using passive markers, generate a huge amount of data. Can topological data analysis methods improve the recognition of movements? Can we use the results of this analysis combined with particular types of neural networks to anticipate the continuation of a movement?

Development of a Kinect Software Tool to Classify Movements during Active Video Gaming.

While it has been established that using full body motion to play active video games results in increased levels of energy expenditure, there is little information on the classification of human movement during active video game play in relationship to fundamental movement skills. The aim of this study was to validate software utilising Kinect sensor motion capture technology to recognise fundamental movement skills (FMS), during active video game play. Two human assessors rated jumping and side-stepping and these assessments were compared to the Kinect Action Recognition Tool (KART), to establish a level of agreement and determine the number of movements completed during five minutes of active video game play, for 43 children (m = 12 years 7 months ± 1 year 6 months). During five minutes of active video game play, inter-rater reliability, when examining the two human raters, was found to be higher for the jump (r = 0.94, p < .01) than the sidestep (r = 0.87, p < .01), although both were excellent. Excellent reliability was also found between human raters and the KART system for the jump (r = 0.84, p, .01) and moderate reliability for sidestep (r = 0.6983, p < .01) during game play, demonstrating that both humans and KART had higher agreement for jumps than sidesteps in the game play condition. The results of the study provide confidence that the Kinect sensor can be used to count the number of jumps and sidestep during five minutes of active video game play with a similar level of accuracy as human raters. However, in contrast to humans, the KART system required a fraction of the time to analyse and tabulate the results.

Mimo space-delay adaptive processing for wide-area indoor human motion classification

This paper develops space-delay adaptive processing (SDAP) for multiple-input, multiple-output (MIMO) arrays with application to classifying human motion in wide-area indoor environments. The proposed adaptive method resolves motion returns using joint spatial-temporal processing on the received signals. Results with real data demonstrate that MIMO SDAP can mitigate ghost motion returns due to indoor multipath scattering and bandwidth limitation for better separation of micro-Doppler signatures from multiple targets. Human activities such as bed turnovers and arm or leg movements are then classified based on micro-Doppler analysis for patient pressure ulcer prevention.

Human motion classification using 2D stick-model matching regression coefficients

Motion estimation methods have been proposed via different approaches, such as silhouette based, model based and image based estimations. However, these methods are highly dependent on the quality of motion data for optimal classification accuracy. Further, because of the complexity of existing algorithms for motion estimation, there are difficulties in interpretation. Hence, the contribution of this work is to model simple human motions for the purpose of recognizing different activity behavior patterns for classification analysis. The model is made up of three body component integrations – Backbone (BB), Upper Body (UB) and Lower Body (LB) – to form a simple 2D human stick figure. Two case studies involving a publicly available video of walking, running and jumping motions as well as experimental captures of Yoga motions are studied. Video motions are simplified into time-step image snapshots, which are later translated into a numeric 2D coordinate system. Initially, the human pelvis is considered the origin of the stick figure. The stick model was drawn by integrating the BB, UB and LB components based on the 2D body joint coordinates. The motion estimation model applies the concept of polynomial fitting to the coordinates data. Computations on the polynomial fitting coefficient deviations at sequential time steps were performed to evaluate the estimation tolerance. A summation of the precedent time-step coordinates with the average deviation metric is used iteratively to estimate the joint coordinates of the stick figure in the subsequent time step to develop the entire motion model. Finally, the developed motion estimation mathematical model was compared to the actual motion phases for classification efficiencies using the Bayes, Lazy, Function, Meta, Misc, Rules and Trees classifiers. Our findings revealed the feasibility of using 2D stick-model matching estimation for human motion classification analysis.

Enlighten Wearable Physiological Monitoring Systems: On-Body RF Characteristics Based Human Motion Classification Using a Support Vector Machine

The real-time health monitoring system is a promising body area network application to enhance the safety of firefighters when they are working in harsh and dangerous environments. Other than monitoring the physiological status of the firefighters, on-body monitoring networks can be also regarded as a candidate solution of motion detection and classification. In this paper, we consider motion classification with features obtained from the on-body radio frequency (RF) channel. Various relevant RF features have been identified and a support vector machine (SVM) has been implemented to facilitate human motion classification. In particular, we distinguish the most frequently appearing human motions of firefighters including standing, walking, running, lying, crawling, climbing, and running upstairs with an average true classification rate of 88.69 percent. Classification performance has been analyzed from three different perspectives including typical classification results, effects of candidate human motions, and effects of on-body sensor locations. We prove that even a subset of available RF features provides an acceptable classification rate, which may result in less computational cost and easier implementation by using our proposed scheme.

Human Motion Classification Using Radio Signal Strength in WBAN

Human Motion Classification Using Radio Signal Strength in WBAN

An Untagged Human Hand Motion Classification by Pyroelectricity

An Untagged Human Hand Motion Classification by Pyroelectricity