Three-dimensional Feature Vector Research Articles

3D+t feature-based descriptor for unsupervised flagellar human sperm beat classification.

Human spermatozoa must swim through the female reproductive tract, where they undergo a series of biochemical and biophysical reactions called capacitation, a necessary step to fertilize the egg. Capacitation promotes changes in the motility pattern. Historically, a two-dimensional analysis has been used to classify sperm motility and clinical fertilization studies. Nevertheless, in a natural environment sperm motility is three-dimensional (3D). Imaging flagella of freely swimming sperm is a difficult task due to their high beating frequency of up to 25 Hz. Very recent studies have described several sperm flagellum 3D beating features (curvature, torsion, asymmetries, etc.). However, up to date, the 3D motility pattern of hyperactivated spermatozoa has not been characterized. The main difficulty in classifying these patterns in 3D is the lack of a ground truth reference since differences in flagellar beat patterns are very difficult to assess visually. Moreover, only around 10-20% of induced to capacitate spermatozoa are truly capacitated, i.e., hyperactivated. We used an image acquisition system that can acquire, segment, and track spermatozoa flagella in 3D+t. In this work, we propose an original three-dimensional feature vector formed by ellipses describing the envelope of the 3D+t spatio-temporal flagellar sperm motility patterns. These features allowed compressing an unlabeled 3D+t dataset to separate hyperactivated cells from others (capacitated from non-capacitated cells) using unsupervised hierarchical clustering. Preliminary results show three main clusters of flagellar motility patterns. The first principal component of these 3D flagella measurements correlated with 2D OpenCASA head determinations as a first approach to validate the unsupervised classification, showing a reasonable correlation coefficient near to 0.7. Clinical relevance- The novelty of this work is defining a 3D+t feature-based descriptor consisting of a set of ellipses enveloping the flagellar motion of human sperm for its unsu-pervised classification. This is a new promising tool to determine the viability of human sperm to fertilize the egg.

Study of degradation of fuel cell stack based on the collected high-dimensional data and clustering algorithms calculations

Accurate perception of the performance degradation of fuel cell is very important to detect its health state. However, inconsistent operating conditions of fuel cell vehicles in the test result in errors in the data. In order to obtain a more credible degradation rate, this study proposes a novel method to classify the experimental data collected under different working conditions into similar operating conditions by using dimensionality reduction and clustering algorithms. Firstly, the experimental data collected from fuel cell vehicles belong to high-dimensional data. Then projecting high-dimensional data into three-dimensional feature vector space via principal component analysis (PCA). The dimension-reduced three-dimensional feature vectors are input into the clustering algorithm, such as K-means and density-based noise application spatial clustering(DBSCAN). According to the clustering results, the fuel cell voltage data with similar operating conditions can be classified. Finally, the selected voltage data can be used to precisely represent the true performance degradation of an on-board fuel cell stack. The results show that the voltage using the K-means algorithm declines the fastest, followed by the DBSCAN algorithm, finally the original data, which indicates that the performance of the fuel cell actually declines faste. Early intervention can prolong its life to the greatest extent.

탄도미사일의 중간단계에서 탄두와 기만체간의 효과적인 변별 연구

For effective detection and tracking of ballistic missiles in the mid-course phase, the discrimination problem between the warhead and decoy needs to be solved. However, because these targets are significantly similar in shape, it is extremely difficult to discriminate them using the representative radar information, such as radar cross-sections and high-resolution range profiles. For this reason, the micro-Doppler information of an echo signal received from a target with micro-motion is usually exploited, in addition to the time-frequency(TF) transform method, to achieve more effective discrimination. However, in conventional methods, only the short time Fourier transform and image moment have been considered as the TF transform and feature extraction methods, respectively, leading to extreme computational complexity. Therefore, in this study, we conducted various simulations to determine the most efficient algorithm from combinations comprising three stages: TF transform, feature extraction, and feature classification. The simulation results obtained using CAD models and an electromagnetic prediction technique reveal that the combination of the Zao-Atlas-Marks distribution with a three-dimensional feature vector and neural network classifier is the most appropriate.

Route Identification Method for On-Ramp Traffic at Adjacent Intersections of Expressway Entrance

To determine the control strategy at intersections adjacent to the expressway on-ramp, a route identification method based on empirical mode decomposition (EMD) and dynamic time warping (DTW) is established. First, the de-noise function of EMD method is applied to eliminate disturbances and extract features and trends of traffic data. Then, DTW is used to measure the similarity of traffic volume time series between intersection approaches and expressway on-ramp. Next, a three-dimensional feature vector is built for every intersection approach traffic flow, including DTW distance, space distance between on-ramp and intersection approach, and intersection traffic volume. Fuzzy C-means clustering method is employed to cluster intersection approaches into classifications and identify critical routes carrying the most traffic to the on-ramp. The traffic data are collected by inductive loops at Xujiahui on-ramp of North and South Viaduct Expressway and surrounding intersections in Shanghai, China. The result shows that the proposed method can achieve route classification among intersections for different time periods in one day, and the clustering result is significantly influenced by three dimensions of traffic flow feature vector. As an illustrative example, micro-simulation models are built with different control strategies. The simulation shows that the coordinated control of critical routes identified by the proposed method has a better performance than coordinated control of arterial roads. Conclusions demonstrated that the proposed route identification method could provide a theoretical basis for the coordinated control of traffic signals among intersections and on-ramp.

Reference region extraction by clustering for the pharmacokinetic analysis of dynamic contrast-enhanced MRI in prostate cancer

PurposeDynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) measures changes in the concentration of an administered contrast agent to quantitatively evaluate blood circulation in a tumor or normal tissues. This method uses a pharmacokinetic analysis based on the time course of a reference region, such as muscle, rather than arterial input function. However, it is difficult to manually define a homogeneous reference region. In the present study, we developed a method for automatic extraction of the reference region using a clustering algorithm based on a time course pattern for DCE-MRI studies of patients with prostate cancer. MethodsTwo feature values related to the shape of the time course were extracted from the time course of all voxels in the DCE-MRI images. Each voxel value of T1-weighted images acquired before administration were also added as anatomical data. Using this three-dimensional feature vector, all voxels were segmented into five clusters by the Gaussian mixture model, and one of these clusters that included the gluteus muscle was selected as the reference region. ResultsEach region of arterial vessel, muscle, and fat was segmented as a different cluster from the tumor and normal tissues in the prostate. In the extracted reference region, other tissue elements including scattered fat and blood vessels were removed from the muscle region. ConclusionsOur proposed method can automatically extract the reference region using the clustering algorithm with three types of features based on the time course pattern and anatomical data. This method may be useful for evaluating tumor circulatory function in DCE-MRI studies.

Median filter detection through streak area analysis

Median filter (MF) is a content preserving nonlinear filter, employed to hide traces of image manipulations, affecting the reliability of manipulation detection techniques. Thus, median filter detection is a major concern for digital image forensics (DIF) experts. The methods used for median filter detection (MFD) are computationally expensive as high dimensionality of feature vectors are employed. This paper proposes an effective method for blind median filter detection based on streaking effect of the median filter. The method offered in the paper is built on experimental observation that the percentage streak area (psa) of an image increases on repetitive median filtering of the image, the rate at which psa increases for median filtered images is different from the rate at which psa increases for unfiltered images. A feature vector based on the observation is extracted from three different image datasets UCID, BOSS and Dresden and feed to Support Vector Machine (SVM) to perform 10-fold cross validation using linear kernel. The results obtained, using a three-dimensional feature vector, demonstrates efficacy of the proposed method.

An automatic feature extraction method and its application in fault diagnosis

The main challenge of fault diagnosis is to extract excellent fault feature, but these methods usually depend on the manpower and prior knowledge. It is desirable to automatically extract useful feature from input data in an unsupervised way. Hence, an automatic feature extraction method is presented in this paper. The proposed method first captures fault feature from the raw vibration signal by sparse filtering. Considering that the learned feature is high-dimensional data which cannot achieve visualization, t-distributed stochastic neighbor embedding (t-SNE) is further selected as the dimensionality reduction tool to map the learned feature into a three-dimensional feature vector. Consequently, the effectiveness of the proposed method is verified using gearbox and bearing experimental datas. The classification results show that the hybrid method of sparse filtering and t-SNE can well extract discriminative information from the raw vibration signal and can clearly distinguish different fault types. Through comparison analysis, it is also validated that the proposed method is superior to the other methods.

The structure function as new integral measure of spatial and temporal properties of multichannel EEG.

The first-order temporal structure functions (SFs), i.e., the first-order statistical moment of absolute increments of scaled multichannel resting state EEG signals in healthy children and teenagers over a wide range of temporal separation (time lags) are computed. Our research shows that the sill level (asymptote) of the SF is mainly defined by a determinant of EEG correlation matrix reflecting the EEG spatial structure. The temporal structure of EEG is found to be characterized by power-law scaling or statistical-scale invariance over time scales less than 0.028 s and at least by two dominant frequencies differing by less than 0.3 Hz. These frequencies define the oscillation behavior of the SF and are mainly distributed within the range of 7.5–12.0 Hz. In this paper, we propose the combined Bessel and exponential model that fits well the empirical SF. It provides a good fit with the mean relative error fitting of 2.8 % over the time lag range of 1 s, using a sampling interval of 4 ms, for all cases under analysis. We also show that the hyper gamma distribution (HGD) fits to the empirical probability density functions (PDFs) of absolute increments of scaled multichannel resting state EEG signals at any given time lag. It means that only two parameters (sample mean of absolute increments and relevant coefficient of variation) may approximately define the empirical PDFs for a given number of channels. A three-dimensional feature vector constructed from the shape and scale parameters of the HGD and the sill level may be used to estimate the closeness of the real EEG to the “random” EEG characterized by the absence of temporal and spatial correlation.

An improved genetic algorithm based on a novel selection strategy for nonlinear programming problems

Genetic algorithm is a heuristic population-based search method that incorporates three primary operators: crossover, mutation and selection. Selection operator plays a crucial role in finding optimal solution for constrained optimization problems. In this paper, an improved genetic algorithm (IGA) based on a novel selection strategy is presented to handle nonlinear programming problems. Each individual in selection process is represented as a three-dimensional feature vector composed of objective function value, the degree of constraints violations and the number of constraints violations. We can distinguish excellent individuals through two indices according to Pareto partial order. Additionally, IGA incorporates a local search ( LS) process into selection operation so as to find feasible solutions located in neighboring areas of some infeasible solutions. Experimental results over a set of benchmark problems demonstrate that proposed IGA has better robustness, effectiveness and stableness than other algorithm reported in literature.

Identity authentication using improved online signature verification method

We present a system for online handwritten signature verification, approaching the problem as a two-class pattern recognition problem. A test signature’s authenticity is established by first aligning it with each reference signature for the claimed user, using dynamic time warping. The distances of the test signature to the nearest, farthest and template reference signatures are normalized by the corresponding mean values obtained from the reference set, to form a three-dimensional feature vector. This feature vector is then classified into one of the two classes (genuine or forgery). A linear classifier used in conjunction with the principal component analysis obtained a 1.4% error rate for a data set of 94 people and 619 test signatures (genuine signatures and skilled forgeries). Our method received the first place at SVC2004 with a 2.8% error rate.

A computational theory for the classification of natural biosonar targets based on a spike code

A computational theory for the classification of natural biosonar targets is developed based on the properties of an example stimulus ensemble. An extensive set of echoes (84 800) from four different foliages was transcribed into a spike code using a parsimonious model (linear filtering, half-wave rectification, thresholding). The spike code is assumed to consist of time differences (interspike intervals) between threshold crossings. Among the elementary interspike intervals flanked by exceedances of adjacent thresholds, a few intervals triggered by disjoint half-cycles of the carrier oscillation stand out in terms of resolvability, visibility across resolution scales and a simple stochastic structure (uncorrelatedness). They are therefore argued to be a stochastic analogue to edges in vision. A three-dimensional feature vector representing these interspike intervals sustained a reliable target classification performance (0.06% classification error) in a sequential probability ratio test, which models sequential processing of echo trains by biological sonar systems. The dimensions of the representation are the first moments of duration and amplitude location of these interspike intervals as well as their number. All three quantities are readily reconciled with known principles of neural signal representation, since they correspond to the centre of gravity of excitation on a neural map and the total amount of excitation.

A computational theory for the classification of natural biosonar targets based on a spike code

A computational theory for the classification of natural biosonar targets is developed based on the properties of an example stimulus ensemble. An extensive set of echoes (84 800) from four different foliages was transcribed into a spike code using a parsimonious model (linear filtering, half-wave rectification, thresholding). The spike code is assumed to consist of time differences (interspike intervals) between threshold crossings. Among the elementary interspike intervals flanked by exceedances of adjacent thresholds, a few intervals triggered by disjoint half-cycles of the carrier oscillation stand out in terms of resolvability, visibility across resolution scales and a simple stochastic structure (uncorrelatedness). They are therefore argued to be a stochastic analogue to edges in vision. A three-dimensional feature vector representing these interspike intervals sustained a reliable target classification performance (0.06% classification error) in a sequential probability ratio test, which models sequential processing of echo trains by biological sonar systems. The dimensions of the representation are the first moments of duration and amplitude location of these interspike intervals as well as their number. All three quantities are readily reconciled with known principles of neural signal representation, since they correspond to the centre of gravity of excitation on a neural map and the total amount of excitation.