Leave-one-subject-out Cross-validation Research Articles

OFF-ApexNet on micro-expression recognition system

When a person attempts to conceal an emotion, the genuine emotion is manifest as a micro-expression. Exploration of automatic facial micro-expression recognition systems is relatively new in the computer vision domain. This is due to the difficulty in implementing optimal feature extraction methods to cope with the subtlety and brief motion characteristics of the expression. Most of the existing approaches extract the subtle facial movements based on hand-crafted features. In this paper, we address the micro-expression recognition task with a convolutional neural network (CNN) architecture, which well integrates the features extracted from each video. We introduce the Optical Flow Features from Apex frame Network (OFF-ApexNet). This is a new feature descriptor that combines the optical flow guided context with the CNN. Firstly, we obtain the location of the apex frame from each video sequence as it portrays the highest intensity of facial motion among all frames. Then, the optical flow information are attained from the apex frame and a reference frame (i.e., onset frame). Finally, the optical flow features are fed into a pre-designed CNN model for further feature enhancement as well as to carry out the expression classification. To evaluate the effectiveness of OFF-ApexNet method, comprehensive evaluations are conducted on three public spontaneous micro-expression datasets (i.e., SMIC, CASME II and SAMM). The promising recognition result suggests that the proposed method can optimally describe the significant micro-expression details. In particular, we report that, in a multi-database with leave-one-subject-out cross-validation (LOSOCV) experimental protocol, the recognition performance reaches 74.60% of recognition accuracy and F-measure of 71.04%. We also note that this is the first work that performs cross-dataset validation on three databases in this domain.

Improving sleep/wake classification with recurrence quantification analysis features

Abstract In this work the method of Recurrence Quantification Analysis (RQA), often used for the analysis of complex dynamic systems, is employed to extract novel features for sleep/wake classification only using cardio-respiratory signals like electrocardiogram (ECG), heart rate (HR) and respiratory effort (RE). A polysomnography data set consisting of 313 full-night recordings is used to evaluate the features. The sleep/wake classification is performed with a classifier based on Linear Discriminant Analysis (LDA) and it is validated in a Leave One Subject Out cross-validation (LOSOCV) scheme. More than 1300 features are extracted from five different cardio-respiratory modalities (ECG, RE, HR, and combinations of ECG + RE and RE + HR). Each modality is processed to obtain 13 basic RQA features, five post-processed versions and, furthermore, three normalizations are applied, leading to a total count of 975 RQA features. From literature, 126 known cardio-respiratory and actigraphy features, normalized with the same procedures thus resulting in 378 distinct features, are used for performance comparison. A feature selection method based on the Mahalanobis distance and the inter-feature correlation is used to determine the most relevant features of both sets, resulting in a set of 158 from the existing features (set A) and 232 from the RQA based features (set B). The pooled Cohen’s kappa coefficient for set A and set B is 0.586 and 0.522, respectively. The combination of both feature sets (set C) improves the kappa value to 0.625. In addition, ROC and PR curves with their corresponding Area Under Curve (AUC) values are computed. It is also shown that derived sleep statistics (sleep efficiency, sleep onset, total sleep time, etc.) deviate less with respect to the ground truth annotations using the additional RQA features compared to the use of literature-based features only. Besides the overall performance, the data set is split into six different but almost equally sized age and sex groups to allow unbiased comparisons among each other. It is shown, that the classification performance decreases with increasing age but is almost independent from sex. Hence, for future work it is suggested to implement the RQA features and also to focus on the improvement of features especially for elderly persons.

Wheeze type classification using non-dyadic wavelet transform based optimal energy ratio technique

Background and objectiveWheezes in pulmonary sounds are anomalies which are often associated with obstructive type of lung diseases. The previous works on wheeze-type classification focused mainly on using fixed time-frequency/scale resolution based on Fourier and wavelet transforms. The main contribution of the proposed method, in which the time-scale resolution can be tuned according to the signal of interest, is to discriminate monophonic and polyphonic wheezes with higher accuracy than previously suggested time and time-frequency/scale based methods. MethodsAn optimal Rational Dilation Wavelet Transform (RADWT) based peak energy ratio (PER) parameter selection method is proposed to discriminate wheeze types. Previously suggested Quartile Frequency Ratios, Mean Crossing Irregularity, Multiple Signal Classification, Mel-frequency Cepstrum and Dyadic Discrete Wavelet Transform approaches are also applied and the superiority of the proposed method is demonstrated in leave-one-out (LOO) and leave-one-subject-out (LOSO) cross validation schemes with support vector machine (SVM), k nearest neighbor (k-NN) and extreme learning machine (ELM) classifiers. ResultsThe results show that the proposed RADWT based method outperforms the state-of-the-art time, frequency, time-frequency and time-scale domain approaches for all classifiers in both LOO and LOSO cross validation settings. The highest accuracy values are obtained as 86% and 82.9% in LOO and LOSO respectively when the proposed PER features are fed into SVM. ConclusionsIt is concluded that time and frequency domain characteristics of wheezes are not steady and hence, tunable time-scale representations are more successful in discriminating polyphonic and monophonic wheezes when compared with conventional fixed resolution representations.

Using Knee Acoustical Emissions for Sensing Joint Health in Patients with Juvenile Idiopathic Arthritis: A Pilot Study.

In this paper, we present a pilot study evaluating novel methods for assessing joint health in patients with Juvenile Idiopathic Arthritis (JIA) using wearable acoustical emission measurements from the knees. Measurements were taken from four control subjects with no known knee injuries, and from four subjects with JIA, before and after treatment. Time and frequency domain features were extracted from the acoustical emission signals and used to compute a knee audio score. The score was used to separate out the two groups of subjects based solely on the sounds their joints produce. It was created using a soft classifier based on gradient boosting trees. The knee audio scores ranged from 0-1 with 0 being a healthy knee and 1 being an involved joint with arthritis. Leave-one-subject-out cross-validation (LOSO-CV) was used to validate the algorithm. The average of the right and left knee audio scores was 0.085±0.099 and 0.89±0.012 for the control group and group with JIA, respectively (p<0.05). The average knee audio score for the subjects with JIA decreased from 0.89±0.012 to 0.25±0.20 following successful treatment (p<0.05). The knee audio score metric successfully distinguished between the control subjects and subjects with JIA. The scores calculated before and after treatment accurately reflected the observed clinical course of the subjects with JIA. After successful treatment, the subjects with JIA were classified as healthy by the algorithm. Knee acoustical emissions provide a novel and cost-effective method for monitoring JIA, and can be used as an objective guide for assessing treatment efficacy.

Eliminating Individual Bias to Improve Stress Detection from Multimodal Physiological Data.

Stress monitoring is important for mental wellbeing and early detection of related disorders. The current work is focused on stress detection from multiple non-invasive physiological signals like Electroencephalogram (EEG), Photoplethysmogram (PPG) and Galvanic Skin Response (GSR). We show that, compared to using only the well known EEG band powers in different frequencies for stress detection, an early fusion with GSR and PPG features shows a significant improvement. Maximum Relevance Minimum Redundancy (mRMR) based feature selection is used to identify the most suitable physiological features correlating with stress. A major contribution of this work lies in eliminating subject-specific bias to improve the classification accuracy. We use self-reported values of Valence, Arousal and Dominance to cluster subjects and build separate classification models specific to clusters. The proposed approach is validated on a publicly available dataset comprising 146 data instances from 10 subjects. The performances of Leave-One- Subject-Out cross validation (LOSOCV) in terms of mean Fscores are 0.61 using EEG features only, 0.64 using early fusion of EEG, GSR and PPG features and 0.69 by applying our clustering technique before fusion and classification.

Validation of non-REM sleep stage decoding from resting state fMRI using linear support vector machines

A growing body of literature suggests that changes in consciousness are reflected in specific connectivity patterns of the brain as obtained from resting state fMRI (rs-fMRI). As simultaneous electroencephalography (EEG) is often unavailable, decoding of potentially confounding sleep patterns from rs-fMRI itself might be useful and improve data interpretation. Linear support vector machine classifiers were trained on combined rs-fMRI/EEG recordings from 25 subjects to separate wakefulness (S0) from non-rapid eye movement (NREM) sleep stages 1 (S1), 2 (S2), slow wave sleep (SW) and all three sleep stages combined (SX). Classifier performance was quantified by a leave-one-subject-out cross-validation (LOSO-CV) and on an independent validation dataset comprising 19 subjects. Results demonstrated excellent performance with areas under the receiver operating characteristics curve (AUCs) close to 1.0 for the discrimination of sleep from wakefulness (S0|SX), S0|S1, S0|S2 and S0|SW, and good to excellent performance for the classification between sleep stages (S1|S2:~0.9; S1|SW:~1.0; S2|SW:~0.8). Application windows of fMRI data from about 70 s were found as minimum to provide reliable classifications. Discrimination patterns pointed to subcortical–cortical connectivity and within-occipital lobe reorganization of connectivity as strongest carriers of discriminative information. In conclusion, we report that functional connectivity analysis allows valid classification of NREM sleep stages.