Real-Time Quality Index to Control Data Loss in Real-Life Cardiac Monitoring Applications.

Wearable cardiac sensors pave the way for advanced cardiac monitoring applications based on heart rate variability (HRV). In real-life settings, heart rate (HR) measurements are subject to motion artifacts that may lead to frequent data loss (missing samples in the HR signal), especially for commercial devices based on photoplethysmography (PPG). The current study had two main goals: (i) to provide a white-box quality index that estimates the amount of missing samples in any piece of HR signal; and (ii) to quantify the impact of data loss on feature extraction in a PPG-based HR signal. This was done by comparing real-life recordings from commercial sensors featuring both PPG (Empatica E4) and ECG (Zephyr BioHarness 3). After an outlier rejection process, our quality index was used to isolate portions of ECG-based HR signals that could be used as benchmark, to validate the output of Empatica E4 at the signal level and at the feature level. Our results showed high accuracy in estimating the mean HR (median error: 3.2%), poor accuracy for short-term HRV features (e.g., median error: 64% for high-frequency power), and mild accuracy for longer-term HRV features (e.g., median error: 25% for low-frequency power). These levels of errors could be reduced by using our quality index to identify time windows with few or no data loss (median errors: 0.0%, 27%, and 6.4% respectively, when no sample was missing). This quality index should be useful in future work to extract reliable cardiac features in real-life measurements, or to conduct a field validation study on wearable cardiac sensors.

A combined mono- and multi-turbine approach for fault indicator synthesis and wind turbine monitoring using SCADA data.

The monitoring of wind turbines using SCADA data has received lately a growing interest from the fault diagnosis community because of the very low cost of these data, which are available in number without the need for any additional sensor. Yet, these data are highly variable due to the turbine constantly changing its operating conditions and to the rapid fluctuations of the environmental conditions (wind speed and direction, air density, turbulence, …). This makes the occurrence of a fault difficult to detect. To address this problem, we propose a multi-level (turbine and farm level) strategy combining a mono- and a multi-turbine approach to create fault indicators insensitive to both operating and environmental conditions. At the turbine level, mono-turbine residuals (i.e. a difference between an actual monitored value and the predicted one) obtained with a normal behavior model expressing the causal relations between variables from the same single turbine and learnt during a normal condition period are calculated for each turbine, so as to get rid of the influence of the operating conditions. At the farm level, the residuals are then compared to a wind farm reference in a multi-turbine approach to obtain fault indicators insensitive to environmental conditions. Indicators for the objective performance evaluation are also proposed to compare wind turbine fault detection methods, which aim at evaluating the cost/benefit of the methods from a production manager's point of view. The performance of the proposed combined mono- and multi-turbine method is evaluated and compared to more classical methods proposed in the literature on a large real data set made of SCADA data recorded on a French wind farm during four years : it is shown than it can improve the fault detection performance when compared to a residual analysis limited at the turbine level only.

Efficient Workload Classification based on Ignored Auditory Probes: A Proof of Concept.

Mental workload is a mental state that is currently one of the main research focuses in neuroergonomics. It can notably be estimated using measurements in electroencephalography (EEG), a method that allows for direct mental state assessment. Auditory probes can be used to elicit event-related potentials (ERPs) that are modulated by workload. Although, some papers do report ERP modulations due to workload using attended or ignored probes, to our knowledge there is no literature regarding effective workload classification based on ignored auditory probes. In this paper, in order to efficiently estimate workload, we advocate for the use of such ignored auditory probes in a single-stimulus paradigm and a signal processing chain that includes a spatial filtering step. The effectiveness of this approach is demonstrated on data acquired from participants that performed the Multi-Attribute Task Battery - II. They carried out this task during two 10-min blocks. Each block corresponded to a workload condition that was pseudorandomly assigned. The easy condition consisted of two monitoring tasks performed in parallel, and the difficult one consisted of those two tasks with an additional plane driving task. Infrequent auditory probes were presented during the tasks and the participants were asked to ignore them. The EEG data were denoised and the probes' ERPs were extracted and spatially filtered using a canonical correlation analysis. Next, binary classification was performed using a Fisher LDA and a fivefold cross-validation procedure. Our method allowed for a very high estimation performance with a classification accuracy above 80% for every participant, and minimal intrusiveness thanks to the use of a single-stimulus paradigm. Therefore, this study paves the way to the efficient use of ERPs for mental state monitoring in close to real-life settings and contributes toward the development of adaptive user interfaces.

A comparison of ERP spatial filtering methods for optimal mental workload estimation.

Mental workload estimation is of crucial interest for user adaptive interfaces and neuroergonomics. Its estimation can be performed using event-related potentials (ERPs) extracted from electroencephalographic recordings (EEG). Several ERP spatial filtering methods have been designed to enhance relevant EEG activity for active brain-computer interfaces. However, to our knowledge, they have not yet been used and compared for mental state monitoring purposes. This paper presents a thorough comparison of three ERP spatial filtering methods: principal component analysis (PCA), canonical correlation analysis (CCA) and the xDAWN algorithm. Those methods are compared in their performance to allow for an accurate classification of mental workload when applied in an otherwise similar processing chain. The data of 20 healthy participants that performed a memory task for 10 minutes each was used for classification. Two levels of mental workload were considered depending on the number of digits participants had to memorize (2/6). The highest performances were obtained using the CCA filtering and the xDAWN algorithm respectively with 98% and 97% of correct classification. Their performances were significantly higher than that obtained using the PCA filtering (88%).

Feature extraction of the first difference of EMG time series for EMG pattern recognition.

This paper demonstrates the utility of a differencing technique to transform surface EMG signals measured during both static and dynamic contractions such that they become more stationary. The technique was evaluated by three stationarity tests consisting of the variation of two statistical properties, i.e., mean and standard deviation, and the reverse arrangements test. As a result of the proposed technique, the first difference of EMG time series became more stationary compared to the original measured signal. Based on this finding, the performance of time-domain features extracted from raw and transformed EMG was investigated via an EMG classification problem (i.e., eight dynamic motions and four EMG channels) on data from 18 subjects. The results show that the classification accuracies of all features extracted from the transformed signals were higher than features extracted from the original signals for six different classifiers including quadratic discriminant analysis. On average, the proposed differencing technique improved classification accuracies by 2-8%.

Mental fatigue and working memory load estimation: interaction and implications for EEG-based passive BCI.

Current mental state monitoring systems, a.k.a. passive brain-computer interfaces (pBCI), allow one to perform a real-time assessment of an operator's cognitive state. In EEG-based systems, typical measurements for workload level assessment are band power estimates in several frequency bands. Mental fatigue, arising from growing time-on-task (TOT), can significantly affect the distribution of these band power features. However, the impact of mental fatigue on workload (WKL) assessment has not yet been evaluated. With this paper we intend to help fill in this lack of knowledge by analyzing the influence of WKL and TOT on EEG band power features, as well as their interaction and its impact on classification performance. Twenty participants underwent an experiment that modulated both their WKL (low/high) and time spent on the task (short/long). Statistical analyses were performed on the EEG signals, behavioral and subjective data. They revealed opposite changes in alpha power distribution between WKL and TOT conditions, as well as a decrease in WKL level discriminability with increasing TOT in both number of statistical differences in band power and classification performance. Implications for pBCI systems and experimental protocol design are discussed.

Probing ECG-based mental state monitoring on short time segments.

Electrocardiography is used to provide features for mental state monitoring systems. There is a need for quick mental state assessment in some applications such as attentive user interfaces. We analyzed how heart rate and heart rate variability features are influenced by working memory load (WKL) and time-on-task (TOT) on very short time segments (5s) with both statistical significance and classification performance results. It is shown that classification of such mental states can be performed on very short time segments and that heart rate is more predictive of TOT level than heart rate variability. However, both features are efficient for WKL level classification. What's more, interesting interaction effects are uncovered: TOT influences WKL level classification either favorably when based on HR, or adversely when based on HRV. Implications for mental state monitoring are discussed.

On-line automatic detection of driver drowsiness using a single electroencephalographic channel.

In this paper, an on-line drowsiness detection algorithm using a single electroencephalographic (EEG) channel is presented. This algorithm is based on a means comparison test to detect changes of the alpha relative power ([8-12]Hz band). The main advantage of the method proposed is that the detection threshold is completely independent of drivers and does not need to be tuned for each person. This algorithm, which works on-line, has been tested on a huge dataset representing 60 hours of driving and give good results with nearly 85% of good detections and 20% of false alarms.

On-line segmentation algorithm for continuously monitored data in intensive care units.

An on-line segmentation algorithm is presented in this paper. It is developed to preprocess data describing the patient's state, sampled at high frequencies in intensive care units, with a further purpose of alarm filtering. The algorithm splits the signal monitored into line segments--continuous or discontinuous--of various lengths and determines on-line when a new segment must be calculated. The delay of detection of a new line segment depends on the importance of the change: the more important the change, the quicker the detection. The linear segments are a correct approximation of the structure of the signal. They emphasise steady-states, level changes and trends occurring on the data. The information returned by the algorithm, which is the time at which the segment begins, its ordinate and its slope, is sufficient to completely reconstruct the filtered signal. This makes the algorithm an interesting tool to provide a processed time history record of the monitored variable. It can also be used to extract on-line information on the signal, such as its trend, in the short or long term.

Evaluation of a device scoring classes of hemorrhagic shock.

In order to evaluate the feasibility of a device scoring classes of hemorrhagic shock, a multivariate analysis of physiological data collected on swine enduring continuous blood loss was conducted. Raw data sampled at up to 500 Hz were first preprocessed and used for features extraction over period of 1 mm. An expert scored all these physiological features, into one of the four classes of hemorrhagic shock: none, compensated, uncompensated and irreversible. A supervised learning of various classifiers was then evaluated over these data. The percentage of misclassification obtained when using a realistic way of estimating error (a leave one -animal- out validation) is about 20% when mean arterial pressure is used, and about 40% when only non invasive features are used. The results are about the same whatever the classifiers used. This evaluation is discussed and a visualization is proposed in order to assess the temporal supervision given by the classifiers.