Emotional and physical stress can cause various health problems. In this paper, we used tissue blood oxygen saturation (StO2), a newly proposed physiological signal, to classify the human stress. We firstly constructed public StO2 database including 42 volunteers subjected two types of During experiment, observed that facial right after be either increased or decreased comparing baseline. investigated feature combinations for classification found average values from left cheek, chin, middle...

In this paper, we propose a method of extracting feature descriptors from discrete spherical images using convolutional neural networks (CNNs). First, captured full-view image is mapped to image. Second, the features-from-accelerated-segment test algorithm used extract points in Finally, an unsupervised CNN obtain patches around each point. experiments, compare these descriptors' performance closest existing state-of-the-art images, oriented FAST and rotated BRIEF (SPHORB), for pairs having...

At present, submarine intelligent path planning is an essential subject due to electronic chart technology and improvement on navigation equipment when sailing has been normalized. Based hydrological environment, a new threat cost calculation method proposed. The distance from the waypoint center length of track in area are added function. pheromone update coalescing differential evolution algorithm aimed solve problem that ant colony local optimization lead improve performance global...

How an image is represented as the input of a convolutional neural network (CNN) important because this directly influences performance CNN. In paper, we investigate representation spherical images by focusing on inclination estimation camera. Unlike other approaches to CNN-based estimation, geodesic-division-based discrete (DSI) that obtained sampling sphere uniformly possible. The CNN single consists five parallelograms flattened from regular icosahedron. To demonstrate advantage proposed...

Stress plays an important role in our daily life. Long-term's psychological stress will lead to serious health as well social problems, it is detect and monitor the its early stage. Most existing detection equipment are contact-type, such wrist strap. However, a real application, working environment, contact-free system bring greater convenience. In this paper, we proposed novel framework for detecting classifying human based on respiratory signals measured remotely by using Kinect sensor...

We present a gaze estimation method for head-mounted full-view egocentric camera that can capture video together with users' cues. While the conventional recording device has two cameras, an eye and scene camera, proposed only single spherical full field of view. To determine point on images, we eye-model-based by means three-dimensional iris projection. First, model is built according to precalibration process eyeball center biological parameters; then, express 3D contour in project it back...

Estimating camera pose is a basic task for computer/robot vision. In this paper we investigated the estimation of an omnidirectional by applying PoseNet [1] to images. An image represented equirectangular image, perspective mosaic representation generated from cubemap, and multi-channel representation. The comparative experiments are carried out these three images effect field view on accuracy estimation. improvement with increase confirmed.

The weakness of vision-based localization methods is that a robot may be confused when similar scenes are observed. In this paper we propose method to cope with problem by using multiple spherical images captured at neighboring places; is, the ambiguity caused visual similarity place decreased observing its places. Concretely, route moves along sampled discrete points which correspond places localized. Then, formulated as classification problem, in (a sampling point route) corresponds class....

Structural and stratigraphic discontinuities developed in subsurface, such as channels, faults fractures, usually contribute to the construction of traps hydrocarbon reservoirs. Seismic attributes are used characterize reservoir features. For instance, coherence is an important attribute measure waveform changes delineate discontinuities, which allows interpreters quickly visualize map complex fault systems. Considering multi-scale characteristic seismic data, spectral decomposition divide...

Aiming at the nonlinear and non- stationary characteristics of vibration signal sequence mechanical equipment, a condition trend prediction method for equipment is proposed based on phase space reconstruction genetic optimization support vector regression (SVR). First all, one- dimensional time series signals are transformed into matrix by use technique, its features selected adaptively. The points imported to SVR model as input predictor trained. Then, adaptive algorithm applied optimize...

Let K he the set of {1,2,....,m}, [x, y] denote [x,x+1,...,y], where 1/spl les/x,y/spl les/m. Given two permutations /spl sigma//sub A/ and B/ a aleph/, A 2-tuple intervals ([x/sub 1/, y/sub 1/], [x/sub 2/, 2/]) is called common if A/([x/sub 1/])=([x/sub 2/]). In this paper, we propose sufficient necessary condition for to be intervals. Based on these conditions, present generic algorithm that finds all permutations.

In order to improve the survival probability of submersible when performing underwater activities, relative error distribution domain is constructed by vector errors and target motion. Analyze elements elements, then establish map with position as center point study mechanism each error. Use submarine heading reference axis, derive synthetic ellipse based on time-varying errors. Based mean square pure orientation solution, uncertain field obtained simulation, which serves basis for avoid...

How to recognize the face expression of a user moving freely is new problem be coped with. In this paper, we propose novel method recognition via profile image captured by head-mounted spherical camera. First, created fisheye dataset. Then, trained 3D convolutional neural network proposed DU Tran [1] for dynamic recognition. As shown in experimental results, achieved as high 72.2% rate.

In this paper we propose a method of localize mobile robot along route by processing spherical images. A camera is mounted on robot. The sampled discrete points which correspond to the places be localized. Then, localization problem formulated as classification problem, in place (a sampling point route) corresponds class. Since image has rotational invariance, content independent rotation. at space, have unique label scenes are different each other. As shown preliminary experimental results,...