IMU2Doppler

The proliferation of sensors powered by state-of-the-art machine learning techniques can now infer context, recognize activities and enable interactions. A key component required to build these automated sensing systems is labeled training data. However, the cost of collecting and labeling new data impedes our ability to deploy new sensors to recognize human activities. We tackle this challenge using domain adaptation i.e., using existing labeled data in a different domain to aid the training of a machine learning model for a new sensor. In this paper, we use off-the-shelf smartwatch IMU datasets to train an activity recognition system for mmWave radar sensor with minimally labeled data. We demonstrate that despite the lack of extensive datasets for mmWave radar, we are able to use our domain adaptation approach to build an activity recognition system that classifies between 10 activities with an accuracy of 70% with only 15 seconds of labeled doppler data. We also present results for a range of available labeled data (10 - 30 seconds) and show that our approach outperforms the baseline in every single scenario. We take our approach a step further and show that multiple IMU datasets can be combined together to act as a single source for our domain adaptation approach. Lastly, we discuss the limitations of our work and how it can impact future research directions.

Tornado Formation and Intensity Prediction Using Polarimetric Radar Estimates of Updraft Area

A sample of 198 supercells are investigated to determine if a radar proxy for the area of the storm midlevel updraft may be a skillful predictor of imminent tornado formation and/or peak tornado intensity. A novel algorithm, a modified version of the Thunderstorm Risk Estimation from Nowcasting Development via Size Sorting (TRENDSS) algorithm is used to estimate the area of the enhanced differential radar reflectivity factor (ZDR) column in Weather Surveillance Radar – 1988 Doppler data; the ZDR column area is used as a proxy for the area of the midlevel updraft. The areas of ZDR columns are compared for 154 tornadic supercells and 44 non-tornadic supercells, including 30+ supercells with tornadoes rated EF1, EF2, and EF3; nine supercells with EF4+ tornadoes also are analyzed. It is found that (i) at the time of their peak 0-1 km azimuthal shear, non-tornadic supercells have consistently small (< 20 km2) ZDR column areas while tornadic cases exhibit much greater variability in areas, and (ii) at the time of tornadogenesis, EF3+ tornadic cases have larger ZDR column areas than tornadic cases rated EF1/2. In addition, all nine violent tornadoes sampled have ZDR column areas > 30 km2 at the time of tornadogenesis. However, only weak positive correlation is found between ZDR column area and both radar-estimated peak tornado intensity and maximum tornado path width. Planned future work focused on mechanisms linking updraft size and tornado formation and intensity is summarized and the use of the modified TRENDSS algorithm, which is immune to ZDR bias and thus ideal for real-time operational use, is emphasized.

A modified Dabestani-Mahan formula estimates a normal pulmonary artery systolic pressure: a single-center retrospective study

Background Pulmonary artery systolic pressure (PASP) is increasingly used as an important datapoint in clinical decision-making and prognostication even in specialties outside of cardiology. Estimation of PASP by Doppler quantification using tricuspid regurgitation (TR) peak velocity is commonly used and correlates well with invasive measurement by right heart catheterization. Further study of transthoracic echocardiogram (TTE) techniques to estimate PASP is needed to provide this datapoint in the absence of sufficient Doppler data for the TR peak velocity method. One technique using right ventricular outflow tract acceleration time (AT) to estimate mean pulmonary artery pressure (MPAP) has been proposed by Dabestani Et al. by the equation MPAP=90-(0.62x AT). Assuming a linear relationship between MPAP and PASP, as suggested by Chemla Et al. by MPAP=(0.61xPASP)+2, a modified formula PASP=145-AT could possibly estimate a normal PASP ≤25 mmHg. Purpose To examine if a modified Dabestani-Mahan formula PASP=145-AT can estimate a normal PASP ≤25 mmHg as calculated by the TR peak velocity method. Methods We queried the electronic medical record at our institution for a sample of 300 patients who had a TTE performed between 2017 and 2020. Each TTE was reviewed and PASP was estimated for each using the TR peak velocity method. A right atrial pressure of 3 mmHg, 8 mmHg, or 15 mmHg was used in the estimation based on inferior vena cava diameter and collapsibility in keeping with the 2015 American Society of Echocardiography guidelines. A short axis view of pulmonary flow using the pulse-waved Doppler sample volume over the transpulmonary valve jet was then reviewed. The time from onset of ejection to peak flow velocity was measured manually as AT in milliseconds using Change Healthcare Cardiology Web Software Package 14.1.1. The measured AT was averaged over three cardiac cycles. Patients with a heart rate between 60 and 100 beats per minute at time of TTE and with sufficient Doppler data to estimate PASP by TR peak velocity and to measure AT were included in a logistic regression analysis. Results 154 patients were included in the statistical analysis. Patients who had a right ventricular outflow tract acceleration time greater than 120 milliseconds, giving a PASP ≤25 mmHg by the modified formula PASP=145-AT, had a 36 times greater odds of having a PASP ≤25 mmHg by the TR peak velocity method (OR=36.0, 95% CI=10.36–125.12, p&lt;0.001). Conclusion(s) Based on a single-center sample, a right ventricular outflow tract acceleration time greater than 120 milliseconds could be used to estimate a normal pulmonary artery systolic pressure less than or equal to 25 mmHg in the absence of sufficient Doppler data for the commonly used TR peak velocity method. FUNDunding Acknowledgement Type of funding sources: None.

A peek into Jupiter’s normal modes from Juno gravity data

&lt;p&gt;As of April 2021, Juno is close to complete its nominal mission, awaiting to enter its extended mission. Thanks to the extremely accurate Doppler data (having an accuracy as low as 10 micron/s at an integration time of 60 s) acquired during close perijove passes in the last 4 years, Juno provided an unprecedented view of Jupiter&amp;#8217;s gravity field, which is crucial to determine its interior structure. In order to recover the gravity field of the planet, the orbits of Juno have to be reconstructed to a very high accuracy. The latest gravity field reconstruction showed hints to a non-static and/or non-axially symmetric field, possibly related to several different phenomena, such as normal modes, localized atmospheric or deeply-rooted dynamics. These tiny phenomena produces a residual signal at a level of few tens of micron/s in Juno Doppler data. To confidently study these tiny unconventional phenomena, the dynamical model of Juno&amp;#8217;s spacecraft have been accurately characterized and possible error sources investigated and ruled out.&lt;/p&gt;&lt;p&gt;The focus of this study is Jupiter&amp;#8217;s normal modes. Our main goal is to assess whether the residuals signatures can be explained by the gravitational disturbances induced by normal modes inside the planet, assuming reasonable physical constraints. Ground-based observations of Jupiter&amp;#8217; normal modes can be used as a guide.&lt;/p&gt;

Precise Orbit Determination of MEX Flyby Phobos Using Simulated Radiometric and Image Data

A navigation camera or topography camera is a standard payload for deep space missions and the image data are normally used for auto-navigation. In this work, we study the potential contribution of image data in precise orbit determination for deep space spacecraft. The Mars Express (MEX) spacecraft has generated extensive Phobos image data during flybys of Phobos, but these data have not been used in precise orbit determination because of the difficulty in employing these image data. Therefore, we did an experiment using simulated image data as the first step for exploring how to use real image data in precise orbit determination of spacecraft. Our results demonstrate that image data can provide stronger constraints on orbit in the tangential and normal directions than Doppler data. When the image data were used in the MEX orbit determination during the MEX Phobos flyby, the orbit determination accuracies in the tangential and normal directions were significantly improved. This work will provide a reference for real image data processing during MEX Phobos flyby to improve MEX orbit accuracy as well as Phobos ephemeris accuracy.

Intracardiac Flow Visualization Using High-Frame Rate Blood Speckle Tracking Echocardiography: Illustrations from infants with Congenital Heart Disease

We report applications of novel high-frame rate blood speckle tracking (BST) echocardiography in a series of infants with congenital heart disease (CHD). BST echocardiography was highly feasible, reproducible, and fast. High-frame rate BST provided complimentary information to conventional color-Doppler data enhancing the visualization and understanding of anomalous blood trajectories (e.g., shunt direction, regurgitant volumes, and stenotic jets) and vortex formation. High-frame rate BST echocardiography is a new, promising imaging tool that may be helpful for deeper understanding of complex CHD physiology.

Using Probe Data Analytics for Assessing Freeway Speed Reductions during Rain Events

Rain impacts roadways such as wet pavement, standing water, decreased visibility, and wind gusts and can lead to hazardous driving conditions. This study investigates the use of high fidelity Doppler data at 1 km spatial and 2-minute temporal resolution in combination with commercial probe speed data on freeways. Segment-based space-mean speeds were used and drops in speeds during rainfall events of 5.5 mm/hour or greater over a one-month period on a section of four to six-lane interstate were assessed. Speed reductions were evaluated as a time series over a 1-hour window with the rain data. Three interpolation methods for estimating rainfall rates were tested and seven metrics were developed for the analysis. The study found sharp drops in speed of more than 40 mph occurred at estimated rainfall rates of 30 mm/hour or greater, but the drops did not become more severe beyond this threshold. The average time of first detected rainfall to impacting speeds was 17 minutes. The bilinear method detected the greatest number of events during the 1-month period, with the most conservative rate of predicted rainfall. The range of rainfall intensities were estimated between 7.5 to 106 mm/hour for the 39 events. This range was much greater than the heavy rainfall categorization at 16 mm/hour in previous studies reported in the literature. The bilinear interpolation method for Doppler data is recommended because it detected the greatest number of events and had the longest rain duration and lowest estimated maximum rainfall out of three methods tested, suggesting the method balanced awareness of the weather conditions around the roadway with isolated, localized rain intensities.

E-wave asymmetry elucidates diastolic ventricular stiffness-relaxation coupling: model-based prediction with in vivo validation

Although diastolic stiffness and relaxation are considered independent chamber properties, the cardio-hemic inertial oscillation that generates E-waves obeys Newton’s law. E-waves vary with heart rate requiring simultaneous change in stiffness and relaxation. By retrospective analysis of human heart-rate varying transmitral Doppler-data, we show that diastolic stiffness and relaxation are coupled and that the coupling manifests through E-wave asymmetry, quantified through a parametrized diastolic filling model-derived dimensionless parameter, which only depends on deceleration time and acceleration time, readily obtainable via standard echocardiography.