A machine learning approach for automated recognition of movement patterns using basic, kinetic and kinematic gait data

This paper investigated application of a machine learning approach (Support vector machine, SVM) for the automatic recognition of gait changes due to ageing using three types of gait measures: basic temporal/spatial, kinetic and kinematic. The gaits of 12 young and 12 elderly participants were recorded and analysed using a synchronized PEAK motion analysis system and a force platform during normal walking. Altogether, 24 gait features describing the three types of gait characteristics were extracted for developing gait recognition models and later testing of generalization performance. Test results indicated an overall accuracy of 91.7% by the SVM in its capacity to distinguish the two gait patterns. The classification ability of the SVM was found to be unaffected across six kernel functions (linear, polynomial, radial basis, exponential radial basis, multi-layer perceptron and spline). Gait recognition rate improved when features were selected from different gait data type. A feature selection algorithm demonstrated that as little as three gait features, one selected from each data type, could effectively distinguish the age groups with 100% accuracy. These results demonstrate considerable potential in applying SVMs in gait classification for many applications.     

PCA-based SVM for automatic recognition of gait patterns

In this technical note, we investigate a combination PCA with SVM to classify gait pattern based on kinetic data. The gait data of 30 young and 30 elderly participants were recorded using a strain gauge force platform during normal walking. The gait features were first extracted from the recorded vertical directional foot- ground reaction forces curve using PCA, and then these extracted features were adopted to develop the SVM gait classifier. The test results indicated that the performance of PCA-based SVM was on average 90% to recognize young- elderly gait patterns, resulting in a markedly improved performance over an artificial neural network-based classifier. The classification ability of the SVM with polynomial and radial basis function kernels was superior to that of the SVM with linear kernel. These results suggest that the proposed technique could provide an effective tool for gait classification in future clinical applications.     

Gender and Age-Related Differences in Bilateral Lower Extremity Mechanics during Treadmill Running

Female runners have a two-fold risk of sustaining certain running-related injuries as compared to their male counterparts. Thus, a comprehensive understanding of the sex-related differences in running kinematics is necessary. However, previous studies have either used discrete time point variables and inferential statistics and/or relatively small subject numbers. Therefore, the first purpose of this study was to use a principal component analysis (PCA) method along with a support vector machine (SVM) classifier to examine the differences in running gait kinematics between female and male runners across a large sample of the running population as well as between two age-specific sub-groups. Bilateral 3-dimensional lower extremity gait kinematic data were collected during treadmill running. Data were analysed on the complete sample (n = 483: female 263, male 220), a younger subject group (n = 56), and an older subject group (n = 51). The PC scores were first sorted by the percentage of variance explained and we also employed a novel approach wherein PCs were sorted based on between-gender statistical effect sizes. An SVM was used to determine if the sex and age conditions were separable and classifiable based on the PCA. Forty PCs explained 84.74% of the variance in the data and an SVM classification accuracy of 86.34% was found between female and male runners. Classification accuracies between genders for younger subjects were higher than a subgroup of older runners. The observed interactions between age and gender suggest these factors must be considered together when trying to create homogenous sub-groups for research purposes.     

Comparison of lower extremity kinematic curves during overground and treadmill running

Researchers conduct gait analyses utilizing both overground and treadmill modes of running. Previous studies comparing these modes analyzed discrete variables. Recently, techniques involving quantitative pattern analysis have assessed kinematic curve similarity in gait. Therefore, the purpose of this study was to compare hip, knee and rearfoot 3-D kinematics between overground and treadmill running using quantitative kinematic curve analysis. Twenty runners ran at 3.35 m/s ± 5% during treadmill and overground conditions while right lower extremity kinematics were recorded. Kinematics of the hip, knee and rearfoot at footstrike and peak were compared using intraclass correlation coefficients. Kinematic curves during stance phase were compared using the trend symmetry method within each subject. The overall average trend symmetry was high, 0.94 (1.0 is perfect symmetry) between running modes. The transverse plane and knee frontal plane exhibited lower similarity (0.86-0.90). Other than a 4.5 degree reduction in rearfoot dorsiflexion at footstrike during treadmill running, all differences were ≤1.5 degrees. 17/18 discrete variables exhibited modest correlations (>0.6) and 8/18 exhibited strong correlations (>0.8). In conclusion, overground and treadmill running kinematic curves were generally similar when averaged across subjects. Although some subjects exhibited differences in transverse plane curves, overall, treadmill running was representative of overground running for most subjects.     

Comparative study of SVM methods combined with voxel selection for object category classification on fMRI data

Background

Support vector machine (SVM) has been widely used as accurate and reliable method to decipher brain patterns from functional MRI (fMRI) data. Previous studies have not found a clear benefit for non-linear (polynomial kernel) SVM versus linear one. Here, a more effective non-linear SVM using radial basis function (RBF) kernel is compared with linear SVM. Different from traditional studies which focused either merely on the evaluation of different types of SVM or the voxel selection methods, we aimed to investigate the overall performance of linear and RBF SVM for fMRI classification together with voxel selection schemes on classification accuracy and time-consuming.

Methodology/Principal Findings

Six different voxel selection methods were employed to decide which voxels of fMRI data would be included in SVM classifiers with linear and RBF kernels in classifying 4-category objects. Then the overall performances of voxel selection and classification methods were compared. Results showed that: (1) Voxel selection had an important impact on the classification accuracy of the classifiers: in a relative low dimensional feature space, RBF SVM outperformed linear SVM significantly; in a relative high dimensional space, linear SVM performed better than its counterpart; (2) Considering the classification accuracy and time-consuming holistically, linear SVM with relative more voxels as features and RBF SVM with small set of voxels (after PCA) could achieve the better accuracy and cost shorter time.

Conclusions/Significance

The present work provides the first empirical result of linear and RBF SVM in classification of fMRI data, combined with voxel selection methods. Based on the findings, if only classification accuracy was concerned, RBF SVM with appropriate small voxels and linear SVM with relative more voxels were two suggested solutions; if users concerned more about the computational time, RBF SVM with relative small set of voxels when part of the principal components were kept as features was a better choice.     

Sampling frequencies and measurement error for linear and temporal gait parameters in primate locomotion

Quantitative analyses of animal motion are increasingly easy to conduct using simple video equipment and relatively inexpensive software packages. With careful use, such analytical tools have the potential to quantify differences in movement between individuals or species and to allow insights into the behavioral consequences of morphological differences between taxa. However, as with any other type of measurement, there are errors associated with kinematic measurements. Because normative kinematic data on human and nonhuman primate locomotion are used to model aspects of gait of fossil hominins, errors in the extant data influence the accuracy of fossil gait reconstructions. The principal goal of this paper is to illustrate the effect of camera speeds (frame rates) on kinematic measurement errors, and to demonstrate how these errors vary with subject size, movement velocity, and sample size. Kinematic data for human walking and running (240 Hz), as well as data for primate quadrupedal walking and running (180 Hz) were used as inputs for a simulation of the measurement errors associated with various linear and temporal kinematic variables. Measurement errors were shown to increase as camera speed, subject body size, and interval duration all decrease, and as movement velocity increases. These results have implications for the methods used to calculate subject velocity and suggest that using a moving marker to measure the linear displacements of the body is preferable to the use of a stationary marker. Finally, while slower camera speeds will always result in higher measurement errors than do faster camera speeds, this effect can be moderated to some extent by collecting sufficiently large samples of data.     

Regression techniques for the prediction of lower limb kinematics

This work presents a novel and extensive investigation of mathematical regression techniques, for the prediction of laboratory-type kinematic measurements during human gait, from wearable measurement devices, such as gyroscopes and accelerometers. Specifically, we examine the hypothesis of predicting the segmental angles of the legs (left and right foot, shank and thighs), from rotational foot velocities and translational foot accelerations. This first investigation is based on kinematic data emulated from motion-capture laboratory equipment. We employ eight established regression algorithms with different properties, ranging from linear methods and neural networks with polynomial support and expanded nonlinearities, to radial basis functions, nearest neighbors and kernel density methods. Data from five gait cycles of eight subjects are used to perform both inter-subject and intra-subject assessments of the prediction capabilities of each algorithm, using cross-validation resampling methods. Regarding the algorithmic suitability to gait prediction, results strongly indicate that nonparametric methods, such as nearest neighbors and kernel density based, are particularly advantageous. Numerical results show high average prediction accuracy (rho = 0.98/0.99, RMS = 5.63 degrees/2.30 degrees, MAD = 4.43 degrees/1.52 degrees for inter/intra-subject testing). The presented work provides a promising and motivating investigation on the feasibility of cost-effective wearable devices used to acquire large volumes of data that are currently collected only from complex laboratory environments.     

Age correction in dementia--matching to a healthy brain

In recent research, many univariate and multivariate approaches have been proposed to improve automatic classification of various dementia syndromes using imaging data. Some of these methods do not provide the possibility to integrate possible confounding variables like age into the statistical evaluation. A similar problem sometimes exists in clinical studies, as it is not always possible to match different clinical groups to each other in all confounding variables, like for example, early-onset (age<65 years) and late-onset (age≥65) patients with Alzheimer's disease (AD). Here, we propose a simple method to control for possible effects of confounding variables such as age prior to statistical evaluation of magnetic resonance imaging (MRI) data using support vector machine classification (SVM) or voxel-based morphometry (VBM). We compare SVM results for the classification of 80 AD patients and 79 healthy control subjects based on MRI data with and without prior age correction. Additionally, we compare VBM results for the comparison of three different groups of AD patients differing in age with the same group of control subjects obtained without including age as covariate, with age as covariate or with prior age correction using the proposed method. SVM classification using the proposed method resulted in higher between-group classification accuracy compared to uncorrected data. Further, applying the proposed age correction substantially improved univariate detection of disease-related grey matter atrophy using VBM in AD patients differing in age from control subjects. The results suggest that the approach proposed in this work is generally suited to control for confounding variables such as age in SVM or VBM analyses. Accordingly, the approach might improve and extend the application of these methods in clinical neurosciences.     

Shoe midsole hardness, sex and age effects on lower extremity kinematics during running

Previous studies investigating the effects of shoe midsole hardness on running kinematics have often used male subjects from within a narrow age range. It is unknown whether shoe midsole hardness has the same kinematic effect on male and female runners as well as runners from different age categories. As sex and age have an effect on running kinematics, it is important to understand if shoe midsole hardness affects the kinematics of these groups in a similar fashion. However, current literature on the effects of sex and age on running kinematics are also limited to a narrow age range distribution in their study population. Therefore, this study tested the influence of three different midsole hardness conditions, sex and age on the lower extremity kinematics during heel-toe running. A comprehensive analysis approach was used to analyze the lower-extremity kinematic gait variables for 93 runners (male and female) aged 16-75 years. Participants ran at 3.33±0.15 m/s on a 30 m-long runway with soft, medium and hard midsoles. A principal component analysis combined with a support vector machine showed that running kinematics based on shoe midsole hardness, sex, and age were separable and classifiable. Shoe midsole hardness demonstrated a subject-independent effect on the kinematics of running. Additionally, it was found that age differences affected the more dominant movement components of running compared to differences due to the sex of a runner.     

Recursive cluster elimination based support vector machine for disease state prediction using resting state functional and effective brain connectivity

Background

Brain state classification has been accomplished using features such as voxel intensities, derived from functional magnetic resonance imaging (fMRI) data, as inputs to efficient classifiers such as support vector machines (SVM) and is based on the spatial localization model of brain function. With the advent of the connectionist model of brain function, features from brain networks may provide increased discriminatory power for brain state classification.

Methodology/Principal Findings

In this study, we introduce a novel framework where in both functional connectivity (FC) based on instantaneous temporal correlation and effective connectivity (EC) based on causal influence in brain networks are used as features in an SVM classifier. In order to derive those features, we adopt a novel approach recently introduced by us called correlation-purged Granger causality (CPGC) in order to obtain both FC and EC from fMRI data simultaneously without the instantaneous correlation contaminating Granger causality. In addition, statistical learning is accelerated and performance accuracy is enhanced by combining recursive cluster elimination (RCE) algorithm with the SVM classifier. We demonstrate the efficacy of the CPGC-based RCE-SVM approach using a specific instance of brain state classification exemplified by disease state prediction. Accordingly, we show that this approach is capable of predicting with 90.3% accuracy whether any given human subject was prenatally exposed to cocaine or not, even when no significant behavioral differences were found between exposed and healthy subjects.

Conclusions/Significance

The framework adopted in this work is quite general in nature with prenatal cocaine exposure being only an illustrative example of the power of this approach. In any brain state classification approach using neuroimaging data, including the directional connectivity information may prove to be a performance enhancer. When brain state classification is used for disease state prediction, our approach may aid the clinicians in performing more accurate diagnosis of diseases in situations where in non-neuroimaging biomarkers may be unable to perform differential diagnosis with certainty.     

Subject-specific and group-based running pattern classification using a single wearable sensor

The objective of this study was to determine whether subject-specific or group-based models provided better classification accuracy to identify changes in biomechanical running gait patterns across different inclination conditions. The classification process was based on measurements from a single wearable sensor using a total of 41,780 strides from eleven recreational runners while running in real-world and uncontrolled environment. Biomechanical variables included pelvic drop, ground contact time, braking, vertical oscillation of pelvis, pelvic rotation, and cadence were recorded during running on three inclination grades: downhill, −2° to −7°; level, −0.2° to +0.2°; and uphill, +2° to +7°. An ensemble and non-linear machine learning algorithm, random forest (RF), was used to classify inclination condition and determine the importance of each of the biomechanical variables. Classification accuracy was determined for subject-specific and group-based RF models. The mean classification accuracy of all subject-specific RF models was 86.29%, while group-based classification accuracy was 76.17%. Braking was identified as the most important variable for all the runners using the group-based model and for most of the runners based on a subject-specific models. In addition, individual runners used different strategies across different inclination conditions and the ranked order of variable importance was unique for each runner. These results demonstrate that subject-specific models can better characterize changes in gait biomechanical patterns compared to a more traditional group-based approach.     

Gait asymmetry: composite scores for mechanical analyses of sprint running

Gait asymmetry analyses are beneficial from clinical, coaching and technology perspectives. Quantifying overall athlete asymmetry would be useful in allowing comparisons between participants, or between asymmetry and other factors, such as sprint running performance. The aim of this study was to develop composite kinematic and kinetic asymmetry scores to quantify athlete asymmetry during maximal speed sprint running. Eight male sprint trained athletes (age 22±5 years, mass 74.0±8.7 kg and stature 1.79±0.07 m) participated in this study. Synchronised sagittal plane kinematic and kinetic data were collected via a CODA motion analysis system, synchronised to two Kistler force plates. Bilateral, lower limb data were collected during the maximal velocity phase of sprint running (velocity=9.05±0.37 ms(-1)). Kinematic and kinetic composite asymmetry scores were developed using the previously established symmetry angle for discrete variables associated with successful sprint performance and comparisons of continuous joint power data. Unlike previous studies quantifying gait asymmetry, the scores incorporated intra-limb variability by excluding variables from the composite scores that did not display significantly larger (p<0.05) asymmetry than intra-limb variability. The variables that contributed to the composite scores and the magnitude of asymmetry observed for each measure varied on an individual participant basis. The new composite scores indicated the inter-participant differences that exist in asymmetry during sprint running and may serve to allow comparisons between overall athlete asymmetry with other important factors such as performance.     

Support vector machine implementations for classification & clustering

BACKGROUND: We describe Support Vector Machine (SVM) applications to classification and clustering of channel current data. SVMs are variational-calculus based methods that are constrained to have structural risk minimization (SRM), i.e., they provide noise tolerant solutions for pattern recognition. The SVM approach encapsulates a significant amount of model-fitting information in the choice of its kernel. In work thus far, novel, information-theoretic, kernels have been successfully employed for notably better performance over standard kernels. Currently there are two approaches for implementing multiclass SVMs. One is called external multi-class that arranges several binary classifiers as a decision tree such that they perform a single-class decision making function, with each leaf corresponding to a unique class. The second approach, namely internal-multiclass, involves solving a single optimization problem corresponding to the entire data set (with multiple hyperplanes). RESULTS: Each SVM approach encapsulates a significant amount of model-fitting information in its choice of kernel. In work thus far, novel, information-theoretic, kernels were successfully employed for notably better performance over standard kernels. Two SVM approaches to multiclass discrimination are described: (1) internal multiclass (with a single optimization), and (2) external multiclass (using an optimized decision tree). We describe benefits of the internal-SVM approach, along with further refinements to the internal-multiclass SVM algorithms that offer significant improvement in training time without sacrificing accuracy. In situations where the data isn't clearly separable, making for poor discrimination, signal clustering is used to provide robust and useful information--to this end, novel, SVM-based clustering methods are also described. As with the classification, there are Internal and External SVM Clustering algorithms, both of which are briefly described.     

Profile-based string kernels for remote homology detection and motif extraction

We introduce novel profile-based string kernels for use with support vector machines (SVMs) for the problems of protein classification and remote homology detection. These kernels use probabilistic profiles, such as those produced by the PSI-BLAST algorithm, to define position-dependent mutation neighborhoods along protein sequences for inexact matching of k-length subsequences ("k-mers") in the data. By use of an efficient data structure, the kernels are fast to compute once the profiles have been obtained. For example, the time needed to run PSI-BLAST in order to build the profiles is significantly longer than both the kernel computation time and the SVM training time. We present remote homology detection experiments based on the SCOP database where we show that profile-based string kernels used with SVM classifiers strongly outperform all recently presented supervised SVM methods. We further examine how to incorporate predicted secondary structure information into the profile kernel to obtain a small but significant performance improvement. We also show how we can use the learned SVM classifier to extract "discriminative sequence motifs"--short regions of the original profile that contribute almost all the weight of the SVM classification score--and show that these discriminative motifs correspond to meaningful structural features in the protein data. The use of PSI-BLAST profiles can be seen as a semi-supervised learning technique, since PSI-BLAST leverages unlabeled data from a large sequence database to build more informative profiles. Recently presented "cluster kernels" give general semi-supervised methods for improving SVM protein classification performance. We show that our profile kernel results also outperform cluster kernels while providing much better scalability to large datasets.     

Evolutionary Optimization of Kernel Weights Improves Protein Complex Comembership Prediction

In recent years, more and more high-throughput data sources useful for protein complex prediction have become available (e.g., gene sequence, mRNA expression, and interactions). The integration of these different data sources can be challenging. Recently, it has been recognized that kernel-based classifiers are well suited for this task. However, the different kernels (data sources) are often combined using equal weights. Although several methods have been developed to optimize kernel weights, no large-scale example of an improvement in classifier performance has been shown yet. In this work, we employ an evolutionary algorithm to determine weights for a larger set of kernels by optimizing a criterion based on the area under the ROC curve. We show that setting the right kernel weights can indeed improve performance. We compare this to the existing kernel weight optimization methods (i.e., (regularized) optimization of the SVM criterion or aligning the kernel with an ideal kernel) and find that these do not result in a significant performance improvement and can even cause a decrease in performance. Results also show that an expert approach of assigning high weights to features with high individual performance is not necessarily the best strategy.     

A cross-validation study to select a classification procedure for clinical diagnosis based on proteomic mass spectrometry

We present an approach to construct a classification rule based on the mass spectrometry data provided by the organizers of the "Classification Competition on Clinical Mass Spectrometry Proteomic Diagnosis Data." Before constructing a classification rule, we attempted to pre-process the data and to select features of the spectra that were likely due to true biological signals (i.e., peptides/proteins). As a result, we selected a set of 92 features. To construct the classification rule, we considered eight methods for selecting a subset of the features, combined with seven classification methods. The performance of the resulting 56 combinations was evaluated by using a cross-validation procedure with 1000 re-sampled data sets. The best result, as indicated by the lowest overall misclassification rate, was obtained by using the whole set of 92 features as the input for a support-vector machine (SVM) with a linear kernel. This method was therefore used to construct the classification rule. For the training data set, the total error rate for the classification rule, as estimated by using leave-one-out cross-validation, was equal to 0.16, with the sensitivity and specificity equal to 0.87 and 0.82, respectively.     

Classification of electrocardiogram signals with support vector machines and genetic algorithms using power spectral features

This paper proposes a new power spectral-based hybrid genetic algorithm-support vector machines (SVMGA) technique to classify five types of electrocardiogram (ECG) beats, namely normal beats and four manifestations of heart arrhythmia. This method employs three modules: a feature extraction module, a classification module and an optimization module. Feature extraction module extracts electrocardiogram's spectral and three timing interval features. Non-parametric power spectral density (PSD) estimation methods are used to extract spectral features. Support vector machine (SVM) is employed as a classifier to recognize the ECG beats. We investigate and compare two such classification approaches. First they are specified experimentally by the trial and error method. In the second technique the approach optimizes the relevant parameters through an intelligent algorithm. These parameters are: Gaussian radial basis function (GRBF) kernel parameter σ and C penalty parameter of SVM classifier. Then their performances in classification of ECG signals are evaluated for eight files obtained from the MIT–BIH arrhythmia database. Classification accuracy of the SVMGA approach proves superior to that of the SVM which has constant and manually extracted parameter.