Motion Recognition Method Research Articles

Retracted: Design of Target Detection and Tracking System for Sports Video

Video motion target tracking has been widely used in the field of sports. Content-based video search has a broad application prospect and attracts more and more researchers’ attention. In view of the shortcomings of moving object detection in motion video at present, a human motion recognition method based on correlation vector machine is proposed. Combined with the improved Gaussian mixture model of moving object detection and tracking method in motion video, moving object detection and tracking are carried out. The design of the system is realized. Firstly, it collects sports videos, considers all kinds of sports objects in sports videos, uses improved background update difference method to detect sports objects, then uses correlation vector machine to recognize human motion, and finally uses AdaBoost classifier to use it in multiple sports videos and sports objects. Carry out follow-up experiments. Results in this way, we can track the sports target in sports video accurately and quickly, and the real-time performance of sports target tracking is better than other tracking methods, which provides a new research tool for tracking sports target in sports video.

Racket Motion Recognition Method Based on Improved Two-Stream Convolution Network

Racket Motion Recognition Method Based on Improved Two-Stream Convolution Network

Lower Limb Motion Recognition Method Based on Improved Wavelet Packet Transform and Unscented Kalman Neural Network

Exoskeleton robot is a typical application to assist the motion of lower limbs. To make the lower extremity exoskeleton more flexible, it is necessary to identify various motion intentions of the lower limbs of the human body. Although more sEMG sensors can be used to identify more lower limb motion intention, with the increase in the number of sensors, more and more data need to be processed. In the process of human motion, the collected sEMG signal is easy to be interfered with noise. To improve the practicality of the lower extremity exoskeleton robot, this paper proposed a wavelet packet transform- (WPT-) based sliding window difference average filtering feature extract algorithm and the unscented Kalman neural network (UKFNN) recognition algorithm. We established an sEMG energy feature model, using a sliding window difference average filtering method to suppress noise interference and extracted stable feature values and using UKF filtering to optimize the neural network weights to improve the adaptability and accuracy of the recognition model. In this paper, we collected the sEMG signals of three muscles to identify six lower limb motion intentions. The average accuracy of 94.83% is proposed in this paper. Experiments show that the algorithm improves the accuracy and anti-interference of motion intention recognition of lower limb sEMG signals. The algorithm is superior to the backpropagation neural network (BPNN) recognition algorithm in the lower limb motion intention recognition and proves the effectiveness, novelty, and reliability of the method in this paper.

A method of motion recognition based on electromyographic signals

In a robot-assisted surgery, a skillful surgeon can perform the operation excellently through flexible wrist motions and rich experience. However, there are little researches about the relationship between the wrist motion and electromyography (EMG) signal of surgeon. To this end, we introduce a classification framework of wrist motion to recognize the common wrist motion of the surgeon based on EMG signals. Generally, surface electromyogram (sEMG) signal has been widely used in prosthetic hand control and medical clinical application. Hence, in this paper, we utilize sEMG signals to evaluate the wrist motions. Eight channels of sEMG signals are captured through a MYO armband from the forearm of the subject. Different kinds of features based on EMG signal, root-mean-square, waveform length, and autoregressive are used to recognize wrist motion through linear discriminant analysis method. We test the impacts on recognition performance from the different sEMG features and different sampling moving window's length. Experimental results have verified the recognition performance of the presented approach. It is validated that the RMS feature can achieve best recognition performance with all different sampling moving window's length in comparison with the WL feature and AR feature.

CSI-HC: A WiFi-Based Indoor Complex Human Motion Recognition Method

WiFi indoor personnel behavior recognition has become the core technology of wireless network perception. However, the existing human behavior recognition methods have great challenges in terms of detection accuracy, intrusion, and complexity of operations. In this paper, we firstly analyze and summarize the existing human motion recognition schemes, and due to the existence of the problems in them, we propose a noninvasive, highly robust complex human motion recognition scheme based on Channel State Information (CSI), that is, CSI-HC, and the traditional Chinese martial art XingYiQuan is verified as a complex motion background. CSI-HC is divided into two phases: offline and online. In the offline phase, the human motion data are collected on the commercial Atheros NIC and a powerful denoising method is constructed by using the Butterworth low-pass filter and wavelet function to filter the outliers in the motion data. Then, through Restricted Boltzmann Machine (RBM) training and classification, we establish offline fingerprint information. In the online phase, SoftMax regression is used to correct the RBM classification to process the motion data collected in real time and the processed real-time data are matched with the offline fingerprint information. On this basis, the recognition of a complex human motion is realized. Finally, through repeated experiments in three classical indoor scenes, the parameter setting and user diversity affecting the accuracy of motion recognition are analyzed and the robustness of CSI-HC is detected. In addition, the performance of the proposed method is compared with that of the existing motion recognition methods. The experimental results show that the average motion recognition rate of CSI-HC in three classic indoor scenes reaches 85.4%, in terms of motion complexity and indoor recognition accuracy. Compared with other algorithms, it has higher stability and robustness.

An action identification method based on FSM and limb dry weight

This article mainly studies the motion recognition method used in the human-centered smart systems. First, we learn feature encoding sequences of the training sets, and then extract subactions from the learned sequences using a statistical model. Based on the hierarchical probabilistic context-free grammar characterization of limb sequences, we generate the grammatical rules of different actions according to the action training sets and characterize the actions and subactions using the finite state machine (FSM). To measure each limb sequence matching degree, we introduce a weight factor of the limb to perform gesture recognition. Also, we can get two sets of the compatibility of recognition probability when identifying the same limb movement sequence. Base on the endings, we derive the feature probability fusion formula of two sets of characteristics of limb movement sequence. Finally, we give the recognition results of two experiments and show the effect of the proposed method on several typical actions. The experimental results show that the proposed action recognition method applied on the same dataset has better recognition accuracy and less time cost over the other methods.

Key Algorithm for Human Motion Recognition in Virtual Reality Video Sequences Based on Hidden Markov Model

This paper provides an in-depth discussion of human motion recognition in Virtual Reality (VR) video sequences through hidden Markov models, which are four steps from VR video acquisition and pre-processing, foreground detection, extraction of human feature parameters, and hidden Markov model human motion recognition. A hybrid Gaussian model was used to build a background model in real-time based on changes in VR video information, and the image was subtracted by the background differential method. The optical flow method was used for foreground detection of the target, and the effects of sparse and dense optical flow were compared to obtain the motion characteristics and optical flow information of the target human body, respectively. Features were extracted for human motion, in terms of common geometric features of the body and optical flow information, respectively. In terms of common geometric information, the width-to-height ratio, perimeter-to-area ratio, center of mass, eccentricity, and feature angle were extracted, respectively. For the optical flow information, optical flow descriptors were constructed using a grid-based approach. And feature fusion was performed for the above two parameters by the k-means method to construct the word pocket model. The hidden Markov model parameters were obtained by using the hidden Markov model for the recognition of human motion and training the human feature parameters for each of the four motions. The recognition of the four common human body movements was realized by the forward-backward algorithm. The test results show that the motion recognition method in this paper has high recognition performance and good anti-interference performance. The time-sequence pooling is used to sort the effective video frame feature sequences to obtain the feature vectors that can represent the dynamic changes of video time sequence; finally, the time-sequence feature vectors are used to train the support vector machine for classification recognition. The recognition accuracy is 65.2% and 89.4% for the HMDB51 and UCF101 datasets, respectively.

Motion Recognition Algorithm in VR Video Based on Dual Feature Fusion and Adaptive Promotion

VR video recognition in complex environment, a motion recognition algorithm based on two-feature fusion and adaptive enhancement is proposed to solve the problems of inaccurate target position, target drift and recognition error caused by the vulnerability to light change, target rotation and occlusion. First, based on the spatio-temporal context (STC) mechanism, image sequence features are extracted through spatio-temporal context relationship and visual system characteristics to reduce the influence of light changes and occlusion on behaviors. Secondly, reliable feature point tracks are obtained through image feature point tracking and background track cutting, and a rich set of action descriptors (AD) are calculated from which local motion information, shape and static appearance information of the track are retained. After that, the principal component analysis operator is introduced to define the double feature fusion rules, and the STC feature and AD feature are combined to form a more accurate and complete feature representation. Finally, adaptive boosting algorithm (ABA) is used to train the classification through the new features obtained and complete the decision judgment of behavior and action. The experimental results show that the proposed algorithm has higher recognition accuracy and robustness compared with the current commonly used motion recognition methods, and can better adapt to complex background and motion changes.

Research on Behaviour Recognition Method for Moving Target Based on Deep Convolutional Neural Network

Aiming at the problem that the average recognition degree of the moving target line is low with the traditional motion target behaviour recognition method, a motion recognition method based on deep convolutional neural network is proposed in this paper. A target model of deep convolutional neural network is constructed and the basic unit of the network is designed by using the model. By setting the unit, the returned unit is calculated into the standard density diagram, and the position of the moving target is determined by the local maximum method to realize the behavior identification of the moving target. The experimental results show that the multi-parameter SICNN256 model is slightly better than other model structures. The average recognition rate and recognition rate of the moving target behavior recognition method based on deep convolutional neural network are higher than those of the traditional method, which proves its effectiveness. Since the frequency of single target is higher than that of multiple recognition and there is no target similarity recognition, similar target error detection cannot be excluded.

Human motion recognition based on Kalman random Forest algorithm and 3D multimedia

A human motion recognition method based on Kalman random forest algorithm model is proposed to improve the accuracy and efficiency of tracking algorithm aiming at the multi-degree-of-freedom problem of traditional human motion recognition and tracking. Firstly, a simple explicit manifold space model of human motion parameters is established. Reconstruction of three-dimensional image features for human motion requires accurate identification of human behavioral features. The human motion process is highly random, and the sudden change in motion amplitude causes a sudden change in the number of shape-based parameters in the three-dimensional feature. The compact low-dimensional model simplifies the difficulty of human motion parameters contour tracking; Secondly, Kalman random forest algorithm is used to estimate the global transformation of action and the different distance measures of body action in each sub-state by using action characteristics through boundary approximation and action sequence updating, so as to reduce the computational complexity of the algorithm; Finally, the single shape change and the dynamic action in motion are modeled, and the dynamic tracking of human motion is realized through the linear combination of human motion shape and style factor decomposition.

Bidirectional estimation between context and motion in motion sequence in which context changes

ABSTRACTWe describe a motion recognition method for reducing recognition errors. The method has a two-layer structure: a lower layer for motion recognition that is affected by the distribution of topics used as context information in the upper layer and an upper layer for the topic distribution that is affected by motion recognition in the bottom layer. We introduce an algorithm for the bottom layer to integrate the motion likelihood calculated using a hidden Markov model and motion appearance probabilities obtained by a topic model. We also use a set of particles to estimate and update contexts on the basis of the result of motion recognition in the bottom layer. The set of particles presents a probabilistic distribution of motion topics, and motion recognition and particle update procedures are performed on each particle. We experimented with two types of sequential motion: a sequence of 33 daily motions and complex motion sequences assuming actual observation. The results showed that the proposed method reduced recognition errors and tracked latent topics better than conventional methods did.

Research on Non-Contact Monitoring System for Human Physiological Signal and Body Movement.

With the rapid increase in the development of miniaturized sensors and embedded devices for vital signs monitoring, personal physiological signal monitoring devices are becoming popular. However, physiological monitoring devices which are worn on the body normally affect the daily activities of people. This problem can be avoided by using a non-contact measuring device like the Doppler radar system, which is more convenient, is private compared to video monitoring, infrared monitoring and other non-contact methods. Additionally real-time physiological monitoring with the Doppler radar system can also obtain signal changes caused by motion changes. As a result, the Doppler radar system not only obtains the information of respiratory and cardiac signals, but also obtains information about body movement. The relevant RF technology could eliminate some interference from body motion with a small amplitude. However, the motion recognition method can also be used to classify related body motion signals. In this paper, a vital sign and body movement monitoring system worked at 2.4 GHz was proposed. It can measure various physiological signs of the human body in a non-contact manner. The accuracy of the non-contact physiological signal monitoring system was analyzed. First, the working distance of the system was tested. Then, the algorithm of mining collective motion signal was classified, and the accuracy was 88%, which could be further improved in the system. In addition, the mean absolute error values of heart rate and respiratory rate were 0.8 beats/min and 3.5 beats/min, respectively, and the reliability of the system was verified by comparing the respiratory waveforms with the contact equipment at different distances.

Motion recognition based on Kinect for human-computer intelligent interaction

At present, human-computer interaction technology has become the focus and hot spot of many scholars. With the deepening of research, facial expression recognition, speech recognition, gesture recognition, face recognition and human motion recognition have become the important content of current human-computer interaction research. In this paper, Microsoft Kinect somatosensory camera is used as the input device of motion, a new method of fast human motion recognition is proposed, and a set of real-time motion recognition and robot control system is designed

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.

Kinect-Based Motion Recognition Tracking Robotic Arm Platform

The development of artificial intelligence technology has promoted the rapid improvement of human-computer interaction. This system uses the Kinect visual image sensor to identify human bone data and complete the recognition of the operator’s movements. Through the filtering process of real-time data by the host computer platform with computer software as the core, the algorithm is programmed to realize the conversion from data to control signals. The system transmits the signal to the lower computer platform with Arduino as the core through the transmission mode of the serial communication, thereby completing the control of the steering gear. In order to verify the feasibility of the theory, the team built a 4-DOF robotic arm control system and completed software development. It can display other functions such as the current bone angle and motion status in real time on the computer operation interface. The experimental data shows that the Kinect-based motion recognition method can effectively complete the tracking of the expected motion and complete the grasping and transfer of the specified objects, which has extremely high operability.

Human motions and emotions recognition inspired by LMA qualities

The purpose of this paper is to describe human motions and emotions that appear on real video images with compact and informative representations. We aimed to recognize expressive motions and analyze the relationship between human body features and emotions. We propose a new descriptor vector for expressive human motions inspired from the Laban movement analysis method (LMA), a descriptive language with an underlying semantics that allows to qualify human motion in its different aspects. The proposed descriptor is fed into a machine learning framework including, random decision forest, multi-layer perceptron and two multiclass support vector machines methods. We evaluated our descriptor first for motion recognition and second for emotion recognition from the analysis of expressive body movements. Preliminary experiments with three public datasets, MSRC-12, MSR Action 3D and UTkinect, showed that our model performs better than many existing motion recognition methods. We also built a dataset composed of 10 control motions (move, turn left, turn right, stop, sit down, wave, dance, introduce yourself, increase velocity, decrease velocity). We tested our descriptor vector and achieved high recognition performance. In the second experimental part, we evaluated our descriptor with a dataset composed of expressive gestures performed with four basic emotions selected from Russell’s Circumplex model of affect (happy, angry, sad and calm). The same machine learning methods were used for human emotions recognition based on expressive motions. A 3D virtual avatar was introduced to reproduce human body motions, and three aspects were analyzed (1) how expressed emotions are classified by humans, (2) how motion descriptor is evaluated by humans, (3) what is the relationship between human emotions and motion features.

A comparative study of motion recognition methods for efficacy assessment of upper limb function

SummaryPhysical disorders are considered to be the most severe disability in patients with hemiplegia after stroke. Currently, most studies have used motion feature extraction methods and machine learning–based methods to evaluate the functional degree of post stroke in hemiplegic patients. This research collected feature data from patients under diverse experimental conditions and then fed them into different machine learning classifiers. However, few studies have compared which classifiers and experimental condition could achieve more precise assessments in a specific condition. In this paper, we compared the accuracy of four different classifiers in a conservative motion recognition method. A motion sensor was used for monitoring the upper limb action, and four conservative machine learning classifiers, which map the features to Fugl‐Meyer scale, were chosen for comparison. Ten post‐stroke hemiplegic‐simulated subjects performed a group of predefined actions, and these motion data were used to generate a group of features reflecting the information of each predefined action. We input the features into four classifiers to generate corresponding classifiers. With the Support Vector Machine classifier, prediction accuracy at 97.79% was achieved in the experiment data, which outperformed previous reports. In conclusion, Support Vector Machines perform better than the other three classifiers in the assessment of the degree of post‐stroke hemiplegics. It is encouraging that results have been generated with the proposed assessment method in this exploratory study.

Rehabilitation motion recognition based on the international biomechanical standards

Since 1990, the International Society of Biomechanics (ISB) has made efforts to provide uniformity to motion description. However, the majority of user studies and technological systems that use motion recognition for rehabilitation, such as virtual and augmented reality (VR and AR), do not follows this standard. Typically, VR and AR techniques recognize motion by comparing human joint positions to a pre-recorded movement, direction, or position relative to an interactive object. Aiming to fill a lack of biomechanical reference on joint motion recognition and analysis for rehabilitation interactive systems, this paper proposes a movement description and recognition method following the ISB standard. The developed method aims to provide computational solutions which are clinically relevant and that uses clinical jargon known to the rehabilitation therapists. The proposed method will also allow motion analysis independent of user's height or its position relative to sensor. We conclude this paper by presenting a comparative analysis of the effectiveness of the Kinect sensors (versions 1 & 2) regarding the recognition of patient exercise motions related to the biomechanical standard. Our solution consists of biomechanical motion classification by identifying in which 3D plane the motion occurs. The motion recognition solution also provides patient posture, motion, and information about incorrect exercise performance. The method was developed by first implementing a body representation and Joint Coordinate System based on ISB standard and then motion angle and biomechanical classifications were performed. Specific methods to measure axial rotations were developed inspired by the traditional goniometry strategy. Also, experiments to evaluate if the method is able to classify the biomechanical motions and, thus, detect whether or not they are being performed correctly were performed using both Kinect v1 and Kinect v2. Motions were performed in correctly and wrongly to also verify false positives. Each classified motion was performed 100 times: 70 correct repetitions (35 at normal speed and 35 fast ones) and 30 wrong ones (out of their respective plane). The motion recognition method showed good efficacy in classifying the biomechanical motions correctly. Most of motions presented great success rate of recognition . higher than 80%, mainly using a 20 degrees tolerance.

Research and Optimization of Dynamic Stochastic Neural Networks in Computer Systems

Computer vision systems are at the forefront of computer research, but there is no complicated computer vision mathematical model. Based on computer system and fuzzy theory, a computer vision service system based on dynamic random neural network is proposed. In addition, a new learning method called Fisher forest is proposed which enables the motion recognition method to use different types or even various numbers of joints for different actions, postures or situations. Fisher forest proposed method can also be used as an effective classifier vector for the present class having a different quality or even to classify different dimensions, and therefore may be used in other applications using the data having different types or dimensions under certain conditions in. After building a more accurate mathematical model, we performed simulation experiments on the model to verify the superiority of the system compared to other traditional computer vision systems.

Human body and limb motion recognition via stacked gated recurrent units network

This study proposes a new radar-based human body and limb motion recognition method that exploited the temporal sequentiality of the motions. A stacked gated recurrent units network (SGRUN) is adopted to extract the dynamic sequential human motion patterns. Since the time-varying Doppler and micro-Doppler signatures can commendably represent such motion patterns, the spectrogram is utilised as the input sequence of the SGRUN. Numerical experiments verify that an SGRUN with two 34-neuron gated recurrent unit layers well classifies and recognises six distinct human body and limb motion types.