We analyse and compare the classification accuracies of six different classifiers for a two-class mental task (mental arithmetic and rest) using functional near-infrared spectroscopy (fNIRS) signals. The signals of the mental arithmetic and rest tasks from the prefrontal cortex region of the brain for seven healthy subjects were acquired using a multichannel continuous-wave imaging system. After removal of the physiological noises, six features were extracted from the oxygenated hemoglobin (HbO) signals. Two- and three-dimensional combinations of those features were used for classification of mental tasks. In the classification, six different modalities, linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), k-nearest neighbour (kNN), the Naïve Bayes approach, support vector machine (SVM), and artificial neural networks (ANN), were utilized. With these classifiers, the average classification accuracies among the seven subjects for the 2- and 3-dimensional combinations of features were 71.6, 90.0, 69.7, 89.8, 89.5, and 91.4% and 79.6, 95.2, 64.5, 94.8, 95.2, and 96.3%, respectively. ANN showed the maximum classification accuracies: 91.4 and 96.3%. In order to validate the results, a statistical significance test was performed, which confirmed that the p values were statistically significant relative to all of the other classifiers (p < 0.005) using HbO signals.

In this study, we determine the optimal feature-combination for classification of functional near-infrared spectroscopy (fNIRS) signals with the best accuracies for development of a two-class brain-computer interface (BCI). Using a multi-channel continuous-wave imaging system, mental arithmetic signals are acquired from the prefrontal cortex of seven healthy subjects. After removing physiological noises, six oxygenated and deoxygenated hemoglobin (HbO and HbR) features-mean, slope, variance, peak, skewness and kurtosis-are calculated. All possible 2-and 3-feature combinations of the calculated features are then used to classify mental arithmetic vs. rest using linear discriminant analysis (LDA). It is found that the combinations containing mean and peak values yielded significantly higher (p < 0.05) classification accuracies for both HbO and HbR than did all of the other combinations, across all of the subjects. These results demonstrate the feasibility of achieving high classification accuracies using mean and peak values of HbO and HbR as features for classification of mental arithmetic vs. rest for a two-class BCI.

BackgroundIn this paper, a novel functional near-infrared spectroscopy (fNIRS)-based brain-computer interface (BCI) framework for control of prosthetic legs and rehabilitation of patients suffering from locomotive disorders is presented.MethodsfNIRS signals are used to initiate and stop the gait cycle, while a nonlinear proportional derivative computed torque controller (PD-CTC) with gravity compensation is used to control the torques of hip and knee joints for minimization of position error. In the present study, the brain signals of walking intention and rest tasks were acquired from the left hemisphere’s primary motor cortex for nine subjects. Thereafter, for removal of motion artifacts and physiological noises, the performances of six different filters (i.e. Kalman, Wiener, Gaussian, hemodynamic response filter (hrf), Band-pass, finite impulse response) were evaluated. Then, six different features were extracted from oxygenated hemoglobin signals, and their different combinations were used for classification. Also, the classification performances of five different classifiers (i.e. k-Nearest Neighbour, quadratic discriminant analysis, linear discriminant analysis (LDA), Naïve Bayes, support vector machine (SVM)) were tested.ResultsThe classification accuracies obtained from SVM using the hrf were significantly higher (p < 0.01) than those of the other classifier/ filter combinations. Those accuracies were 77.5, 72.5, 68.3, 74.2, 73.3, 80.8, 65, 76.7, and 86.7% for the nine subjects, respectively.ConclusionThe control commands generated using the classifiers initiated and stopped the gait cycle of the prosthetic leg, the knee and hip torques of which were controlled using the PD-CTC to minimize the position error. The proposed scheme can be effectively used for neurofeedback training and rehabilitation of lower-limb amputees and paralyzed patients.

No abstract

Objective. In this paper, a novel methodology for feature extraction to enhance classification accuracy of functional near-infrared spectroscopy (fNIRS)-based two-class and three-class brain–computer interface (BCI) is presented. Approach. Novel features are extracted using vector-based phase analysis method. Changes in oxygenated Δ H b O and de-oxygenated ( Δ H b R ) haemoglobin are used to calculate four novel features: change in cerebral blood volume ( Δ C B V ), change in cerebral oxygen exchange ( Δ C O E ), vector magnitude (|L|) and angle (k). Δ C B V is the sum and Δ C O E is difference of Δ H b O and Δ H b R , whereas |L| is magnitude and k is angle of vector. fNIRS signals of seven healthy subjects, corresponding to left-hand index finger tapping (LFT), right-hand index finger tapping (RFT) and rest are acquired from motor cortex using multi-channel continuous-wave imaging system. After removing physiological and instrumental noises from the acquired signals, the four novel features are calculated. For validation, conventional temporal, spatial and spatiotemporal features; mean, peak, slope, variance, kurtosis and skewness are also calculated using Δ H b O and Δ H b R . All possible two-feature and three-feature combinations of the novel and conventional features are then used to classify two-class (LFT vs RFT) and three-class (LFT vs RFT vs rest) fNIRS-BCI using linear discriminant analysis. Main results. Results demonstrate that combination of four novel features yields significantly higher average classification accuracies of 98.7 ± 1.0% and 85.4 ± 1.4% as compared to 68.7 ± 6.9% and 53.6 ± 10.6% using conventional features for two-class and three-class problem, respectively. Validation of proposed method on an open access database containing RFT, LFT and dominant side foot tapping tasks for 30 subjects also shows improvement in average classification accuracies for two-class and three-class fNIRS-BCIs. Significance. This study provides a step forward in improving the classification accuracies of state-of-the-art fNIRS-BCIs by showing significant improvement in classification accuracies of two-class and three-class fNIRS-BCIs using novel features extracted by vector-based phase analysis.

In this paper, a novel methodology for enhanced classification of functional near-infrared spectroscopy (fNIRS) signals utilizable in a two-class [motor imagery (MI) and rest; mental rotation (MR) and rest] brain–computer interface (BCI) is presented. First, fNIRS signals corresponding to MI and MR are acquired from the motor and prefrontal cortex, respectively, afterward, filtered to remove physiological noises. Then, the signals are modeled using the general linear model, the coefficients of which are adaptively estimated using the least squares technique. Subsequently, multiple feature combinations of estimated coefficients were used for classification. The best classification accuracies achieved for five subjects, for MI versus rest are 79.5, 83.7, 82.6, 81.4, and 84.1% whereas those for MR versus rest are 85.5, 85.2, 87.8, 83.7, and 84.8%, respectively, using support vector machine. These results are compared with the best classification accuracies obtained using the conventional hemodynamic response. By means of the proposed methodology, the average classification accuracy obtained was significantly higher (p < 0.05). These results serve to demonstrate the feasibility of developing a high-classification-performance fNIRS-BCI.

Smart homes may be beneficial for people of all ages, but this is especially true for those with care needs, such as the elderly. To assist, monitor for emergencies, and provide companionship for the elderly, a substantial amount of research on human activity recognition systems has been conducted. Several algorithms for activity recognition and prediction of future events have been reported in the scientific literature. However, the majority of published research does not address privacy concerns or employ a variety of ambient sensors.The objective of this thesis is to contribute to the progress in research relevant to activity recognition systems that use sensors that collect less privacy-related information. The following tasks are included in the work: assessment of sensors while keeping privacy concerns in mind, selection of cutting-edge classification methods, and how to fuse the data from multiple sensors. This thesis contributes to making progress on systems for analyzing human activity and state-or vital signs-for application in a mobile robot.This dissertation examines two topics. First, it examines the privacy concerns associated with having a robot in the home. On a robot, an ultra-wideband (UWB) radar-based sensor and an RGB camera (for ground truth) were installed. An actigraphy device was also worn by the users for heart rate monitoring. The UWB sensor was selected to maintain privacy while monitoring human activities.Considering different ways to represent data from a single sensor is the second topic under investigation. That is, how data from multiple representations can be combined. For this purpose, we investigate various data representations from a single sensor's data and analysis using cutting-edge deep learning algorithms.The contributions provide considerations for equipping a mobile home robot with activity recognition abilities while reducing the amount of privacy-sensitive sensor data. The work also concerns examining the potential privacy restrictions that must be established for the analyzing systems. The thesis contains new methods for combining data from multiple information sources. To achieve our objective, convolutional neural networks and recurrent neural networks were applied and validated using conventional methods.The conclusion of the thesis is that we can achieve good accuracy with limited sensors while maintaining privacy. It is, however, likely adequate for assisting healthcare personnel and caregivers in their work by indicating current activity status and measuring activity levels, providing alerts about abnormal activities. The results can hopefully contribute to older people being able to live alone in their homes with a larger chance of any unwanted events being quickly detected and notified to the caregivers and providers.iii PrefaceThis thesis is submitted in partial fulfillment of the requirements for the degree of Philosophiae Doctor at the University of Oslo.The research presented here was conducted at the Robotics and Intelligent Systems group at the Departmen...

Teaching machines to learn patterns in data is very common these days, and it has a broad spectrum of applications everywhere. Sensors like smart-watches are getting more functionality each year, and more and more people buy them. Passing data from the watches, for example, activity or heart rate to machine learning algorithms, can generate significant results within many fields. Mental health is an example of a field where computer-generated predictions can be helpful to gain knowledge about patients. For example, machine learning can help predict that someone has a specific type of mental disorder. In this thesis, we present applied machine learning to detect depression. The dataset (collected for another study about behavioral patterns in schizophrenia vs. major depression) contains motor activity measurements for each minute in the measured period for each participant. Three machine learning models are trained to fit time-sliced segments of these measurements. The first model classifies participants into condition group (depressed) and control group (non-depressed). We trained another model to classify the depression level of participants (normal, mild or moderate). Finally, we trained a model that predicts MADRS scores. We evaluate the performance of classification models using leave one participant out validation as a technique, in which we achieved an average F1-score of 0.70 for detecting control/condition group and 0.30 for detecting the depression levels. The MADRS score prediction resulted in a mean squared error of approximately 4.0. These performance scores are promising, but not good enough to be used in the real world. However, not much more work is needed for the first model if we apply it to a dataset with more participants.