Motoki Sakai

Feasibility Study on Blood Pressure Estimations from Voice Spectrum Analysis

Recently, preventive healthcare has attracted increased attention. Because of a demonstrable mortality rate, hypertension is a disease that should be prevented before it becomes severe. To decrease the mortality rate for hypertension, it is desirable for people to measure their own blood pressure (BP) on a routine basis. At present, many ambulatory BP gauges are available. Yet, it is unlikely that people will carry around an ambulatory BP gauge to measure their BP, even among those with a high risk of illness. Such methods are therefore ineffective for preventive healthcare. Therefore, it would be preferable to measure information related to BP easily and without a dedicated gauge. In this research, a BP estimation method is proposed using voicespectrum analysis. If BP estimations from the voice spectrum are accurate, we may be able to measure BP with a smartphone’s voice recorder. To evaluate the feasibility of such BP estimations, two subjects’ BPs and voice data were measured, and the correlation coefficients were examined. Results showed that both diastolic and systolic BPs and the spectral component of the voice were not non-correlated ( > |0.6|) at specific frequency bands. To estimate BP levels, a support-vector machine was proposed, and the correlation coefficients between measured and estimated BPs exceeded 0.9. However, individual differences in the voice spectrum were not adequately addressed. In future research, individual differences will be investigated in a study involving more subjects. General Terms Data mining, Biomedical engineering

Development of Lead System for ECG-Derived Respiration Aimed at Detection of Obstructive Sleep Apnea Syndrome

Obstructive sleep apnea syndrome is the most common type of sleep apnea, characterized by repetitive pauses in breathing during sleep. Recent studies have investigated screening methods based only on an electrocardiogram (ECG). Generally, in ECG-based screening, the ECG-derived respiration (EDR) is often used, which is caused by the variance in the cardiac electrical axis due to the chest movement associated with the respiration itself. This method might be effective for diagnosing the sleep apnea severity, called the apnea-hypopnea index, which is defined by the duration and occurrence rate of apnea episodes. However, conventional ECG lead systems are not necessarily optimized for this purpose. In this study, nine bipolar electrodes located across the entire ventral thoracic region were devised, based on the conventional lead system, to effectively measure thoracic breathing. To evaluate the most effective electrode placements, two eupneic and three apneic tasks with nine electrodes were conducted, and EDRs were calculated. Then, the respiratory rates were estimated from the EDRs, and the eupnea and apnea groups were classified using features of the EDRs. Consequently, it was found that three electrodes located at the lower thoracic region yielded accurate estimations of the respiratory rate and discrimination rate.

A novel information dissemination system for vehicle-to-RSU communication networks

We design a novel information dissemination system for VANET which is capable to selectively provide local information for drivers as much as possible while meets a certain condition of transmission capacity.

Nonlinear State-Space Projection Based Method to Acquire EEG and ECG Components Using a Single Electrode

In some diagnoses, such as the polysomnography, simultaneous measurement of the electroencephalogram (EEG) and the electrocardiogram (ECG) is often required. It would be more efficient if both the EEG and ECG could be obtained simultaneously by using a single measurement. In this paper, we introduce a nonlinear state-space projection-based technique to extract the EEG and ECG components from an EEG signal measured with a non-cephalic reference (NCR) that guarantees accurate detection of R waves in the EEG measurement. Evaluation of the method using simulated data showed that the improved normalized power spectrum in alpha, beta (13-30 Hz), and theta bands were accurate. In an accrual EEG, measured using the NCR electrode, it was confirmed that the frequency components of the extracted EEG were accurate, and no spikes that could be attributed to the ECG component were observed in the resultant EEG signal.

Kernel Nonnegative Matrix Factorization with Constraint Increasing the Discriminability of Two Classes for the EEG Feature Extraction

Nonnegative matrix factorization (NMF) is an algorithm for blind source separation. It has been reported that the use of kernel NMF (KNMF) is a particularly feasible way to extract the features of a motor-imagery related EEG spectrum, which is often used in brain-computer interfaces (BCI). A BCI system enables users to control electrical devices without their hands or feet, and often requests to tell users intention from motor-imagery related EEG features. In other words, a classification of the EEG signals reflecting the users intentions is required. In this research, a constraint is placed on the KNMF to increase the discriminability between two classes, widening the difference between their spectral EEG energies. To evaluate the proposed method, the IDIAP database, which contains the motor-imagery related EEG spectrum of three subjects, was adopted for the discrimination between two classes. As a result, the classification accuracy when using the proposed constraint was approximately 78% on average, which is 4% higher than that obtained by KNMF without a constraint.

Separation of EEG and ECG components based on wavelet shrinkage and variable cosine window

During ambulatory monitoring, it is sometimes required to record an electroencephalogram (EEG) and an electrocardiogram (ECG) simultaneously. It would be ideal if both EEG and ECG could be obtained with one measurement. Here, we introduce an algorithm that combines the wavelet shrinkage and variable cosine window operation to separate the EEG and ECG components from an EEG signal recorded with a noncephalic reference (NCR). Evaluation using simulated data and actual measured data showed that accurate frequency analysis of EEG and an R-R detection-based heart rate analysis were feasible with our proposed algorithm, which improved the signal-averaging based algorithm so that ECG components containing ectopic beats can be applied.

Case study on analysis of vocal frequency to estimate blood pressure

Preventive healthcare is a precaution for a premorbid state. Because of its demonstrable mortality rate, hypertension is a disease that should be prevented before it becomes severe. To decrease the mortality rate for hypertension, it is desirable for people to measure their own blood pressure (BP) on a routine basis. At present, many ambulatory BP gauges are available. However, people are unlikely to carry around a gauge to measure their BP-even those with a high risk of illness. Such methods are ineffective for preventive healthcare. Therefore, it would be preferable to measure factors related easily to BP without a dedicated gauge. In this research, a BP estimation method is proposed using voice spectrum analysis. If BP estimations from the voice spectrum are accurate, we can possibly measure BP using the voice recorder of a smartphone. To evaluate the feasibility of BP estimation, measurement experiments were conducted with two subjects. In the experiments, 60 sets of BP and voice data were measured from each subject, and the correlation coefficients were examined. Results showed that both diastolic and systolic BPs were not uncorrelated with the spectral component of the voice (> |0.6|) at specific frequency bands. To estimate the BP value from voice data, a relationship between BP and voice spectrum was modeled. In the proposed model, BP can be expressed as the polynomial function in terms of the voice spectrum. As a result, correlation coefficients between measured and estimated BPs were approximately 0.9, and average BP estimation errors were small to a certain extent.

Nonnegative Matrix Factorization Algorithms using a Constraint to Increase the Discriminability of Two Classes for EEG Feature Extraction

onnegative matrix factorization (NMF) is an algorithm used for blind source separation. It has been reported that NMF algorithms can be utilized as an effective means to extract features from a motor-imagery related EEG spectrum, which is often used in brain-computer interfaces (BCI). BCI systems enable users to control electrical devices without moving their body parts, and are often tasked with interpreting a users intentions through motor-imagery related EEG features. In other words, they require EEG signal classification in order to reflect user intentions. In this study, constraints are placed on NMF and kernel NMF (KNMF) algorithms to increase the discriminability between two classes by increasing the energy difference between their potential sources in a spectral EEG signal. To evaluate the proposed algorithms, the IDIAP database, which contains the motor-imagery related EEG spectrum of three subjects, was adopted to test the discrimination between two classes. Using the database, the classification accuracy of the proposed constraint was 75%, which was 7% higher than what was obtained through NMF without a constraint. Similarly, the classification accuracy of KNMF with the proposed constraint was also 4% higher than that of KNMF without a constraint, and reached 78%.