Goal: P300 spellers are among the most popular types of brain-computer interfaces (BCIs) and extremely useful assistive devices that enable severely disabled patients to communicate. However, speller performances should be further improved translate laboratory designs into practical applications. We aimed design a new paradigm could evoke higher event-related potentials (ERPs) than traditional spellers, thus improving performance BCI systems. Methods: proposed based on three-dimensional...

Extracting and fusing part features have become the key of fined-grained image recognition. Recently, Non-local (NL) module has shown excellent improvement in However, it lacks mechanism to model interactions between multi-scale features, which is vital for fine-grained In this paper, we propose a novel cross-layer non-local (CNL) associate receptive fields by two operations. First, CNL computes correlations query layer all response layers. Second, are weighted according added features. Due...

Incomplete optical distortion correction in VR HMDs leads to spatial dynamic distortion, which is a potential cause of VIMS. A perception experiment designed for the investigation with three levels, subjective SSQ, five-scale VIMS level rating, and objective postural instability adopted as evaluation metrics. The results show that factor has significant effect on all metrics increments (p<0.05). As drops off, symptoms decrease. study highlights importance perfect eliminating aggravation effect.

Abstract Although several crosstalk definitions have been proposed to indicate the level of in stereoscopic displays, they do not necessarily correlate with perceived crosstalk. Based on two perception experiments, each a different approach, we studied correspondence between 10 objective equations and The first experiment, which was evaluated, demonstrates that based Weber's law (Δ L / ) show highest correlation for most critical case, is, center circle an outer disc. A second visibility...

Rehearsal, retraining on a stored small data subset of old tasks, has been proven effective in solving catastrophic forgetting continual learning. However, due to the sampled may have large bias towards original dataset, them is susceptible driving domain drift tasks feature space, resulting forgetting. In this paper, we focus tackling problem with centroid distance distillation. First, propose caching mechanism for sampling points based constructed centroids reduce sample rehearsal. Then,...

Abstract Although numerous potential causes may lead to visual discomfort when viewing content on three‐dimensional (3D) displays, vergence–accommodation conflict is a particular cause of binocular parallax‐based stereoscopic and it unavoidable. Based the study 3D attention, we proposed novel depth adjustment method improve comfort enhance perceived naturalness. The combined image saliency specific condition establish model for computing optimum zero‐disparity plane image. results perception...

Currently the individual variability in Cyclopean Eye's position is not considered when presenting virtual reality content. We proposed a novel perceptual space model for content based on location of Eye. The methodology may improve immersive viewing experience and interaction accuracy between system user.

Binocular tone mapping aims to improve the overall quality of image erceived by human eye, and has been continuously optimized since it was proposed. Based on visual perception experiment, a series experiments are designed for binocular ability eyes. On this basis, based non‐reference algorithm is The more streamlined closer framework, adaptive in practice.

Abstract The sensory rearrangement theory announced that a mismatch between the current visual inputs and exposure history might trigger or aggravate visually induced motion sickness (VIMS). However, this has not been explicitly investigated through perception experiments. This study investigates effect of mismatches 3D content acquisition on VIMS. Two types camera field view eye were compared with match type, are conditions configured to be larger smaller than view. A experiment was...

In order to avoid toxic gas poisoning incident, explosion and improve people's quality of life, this project designed the wireless low power detector based on ADuCM360 with sensor technology transmission technology. The system consists management module, signal conditioning module communication module. Analysis experimental results show that can monitor store concentration, temperature humidity in room for a long time. You will be able understand detection gases without necessity visiting...

The three‐dimensional endoscope system is widely used for surgery and relative training of novice. geometric distortions will affect the accuracy size distance estimation, also may be one important factor inducing visual discomfort. Based on model image capture display, are analyzed. results provide some theoretical guidance appropriate utilization system.

Existing class-incremental lifelong learning studies only the data is with single-label, which limits its adaptation to multi-label data. This paper Lifelong Multi-Label (LML) classification, builds an online classifier in a sequential classification stream. Training on Partial Labels LML may result more serious Catastrophic Forgetting old classes. To solve problem, study proposes Augmented Graph Convolutional Network (AGCN) built Correlation Matrix (ACM) across partial-label tasks. The...

Text-guided image editing (TIE) seeks to manipulate images by the guidance of language. However, existing TIE methods always overlook target-irrelevant pixels and may make background discolored, distorted, or partially disappear. To overcome this problem, we propose a novel method named FocusGAN, which edits text-relevant precisely as well keeps invariant. Specifically, build network two stages. In each stage, first construct channel-wise subject focusing attention generator focus on...