Natural Video Research Articles

Understanding how virtual reality forest experience promote physiological and psychological health for patients undergoing hemodialysis.

Growing research has found that exposing patients to forest landscapes through technology improves their health. However, studies on the effects of nature therapy sessions on particularly vulnerable groups that require frequent regular treatment such as patients of chronic kidney disease (CKD) is lacking. This study aims to understand the effects of watching panoramic videos of forest walks through virtual reality (VR) headsets on patients. We also examine the effects of the frequency of virtual exposure to forests on the physiological and psychological parameters of patients undergoing dialysis. Twenty-four dialysis patients with a mean age of 65.11 year underwent a virtual nature intervention over the course of 3 weeks. The intervention consists of 3 numbers of 6-min 360-degree nature videos delivered using VR headsets. We measured heart rate variability (HRV) and heart rate (HR) using continuous electrocardiographic (ECG) monitoring and assessed their emotional states using a questionnaire survey and in-depth interviews. The results showed that the initial 6 min of watching of the panoramic videos through VR headsets resulted in significantly reduced HR and sympathetic nerve activity and increased positive emotional responses. However, repeated VR exposure did not further improve the physiological health of the respondents. Considering these results, the use of VR headsets to watch 6-min nature videos indoors can be used to enhance the positive psychological responses among dialysis and bed-ridden patients. As this study is preliminary, future random controlled trials are needed to compare and determine the best dose, content, and delivery methods of such an intervention.

Effects on Wellbeing of Exposure to Dog Videos Before a Stressor

ABSTRACT Animal-assisted intervention (AAI) has been used as a means of stress relief in clinical and general settings; however, animals are not always allowed in certain spaces. Adapting AAI to video or virtual mediums could improve accessibility and is temporally relevant given the recent shift to online interventions. The current study explored: (1) whether an active video (dog or nature) watched before a stressor would improve wellbeing more than tranquil videos; (2) whether exposure to a dog video improves wellbeing more than a nature video; and (3) whether exposure to either a dog or nature video improves outcomes more than exposure to a control video. One hundred and seven undergraduates were randomly assigned to watch one of five videos (active dog, tranquil dog, active nature, tranquil nature, and control) for 3 minutes and then complete a 3-minute stress task. Subjective (anxiety, stress, happiness, relaxation, positive affect, and negative affect) and physiological (blood pressure and heart rate) outcomes were collected at baseline, video, stressor, and recovery time points. Results showed that the activity level of the dog in the video did not influence outcomes. However, relative to the control group, the dog-video condition showed decreases in stress from baseline to video and a smaller decrease in stress from stressor to recovery. Additionally, relative to the nature-video condition, the dog-video condition showed a slightly higher increase in happiness scores from baseline to video. Lastly, relative to the control group, the nature-video condition showed increased relaxation scores from baseline to video and a larger decrease in relaxation scores from video to stressor. This research may inform the development of alternate modes of AAIs.

Physiologically-inspired neural model for social interaction recognition from abstract and naturalistic videos

Physiologically-inspired neural model for social interaction recognition from abstract and naturalistic videos

IdentityMask: Deep Motion Flow Guided Reversible Face Video De-Identification

Unprecedented video collection and sharing have exacerbated privacy concerns and led to increasing interest in privacy-preserving tools. A satisfactory video de-identification tool should be able to remove sensitive identity information from face videos while maintaining useful information for other identity-agnostic tasks. Meanwhile, it is necessary to allow the authority to inspect real identity when abnormal events are detected. Existing methods only focus on the study of de-identification, and lack the desired recovery ability when granting permissions. Furthermore, they all process the videos frame by frame, which hardly benefit from motion and inter-frame information. In this paper, we propose a modular architecture for reversible face video de-identification, called IdentityMask, which leverages deep motion flow to avoid per-frame evaluation. Our framework consists of two processes: the de-identification process provides a protective mask for identity information, while the recovery process can remove the protective mask if and only if the right key is provided. To this end, a Protection Module and a Recovery Module are built as two major functional modules, both based on an identity disentanglement network and guided by a crucial Motion Flow Module. An Affine Transformation Module provides simple but reliable assistance. Extensive experiments on a diverse natural video dataset (gender, ethnicity, age, etc.) demonstrate the effectiveness of the proposed framework for reversible face video de-identification.

The effect of combined Action Observation Therapy with eccentric exercises in the treatment of mid-portion Achilles-tendinopathy: a feasibility pilot randomised controlled trial

BackgroundMid-portion Achilles Tendinopathy (AT) is a common musculoskeletal condition with varying rehabilitation success rates. Despite the prevalence of this condition, a considerable proportion of individuals experience persisting pain and functional deficits. Current treatment approaches bias the biomedical model which emphasises physically treating and loading the tendon. Overall, there is a lack of consideration for the central nervous system that is commonly implicated in chronic injuries. The aim of this pilot study was to explore the feasibility of combining Action Observation Therapy (AOT), a treatment technique which targets central changes and influences motor learning, with eccentric exercises in the treatment of mid-portion AT. AOT involves the observation of movements and is commonly followed by the physical performance of these same movements.MethodologyThis was a double-blinded randomised controlled pilot feasibility study. All participants underwent the 12-week Alfredson eccentric training protocol. The intervention group watched videos of the exercises prior to performing these exercises, whilst the control group watched nature videos before performing the same exercises. Study feasibility was the primary outcome measure, with the Victorian Institute of Sports Assessment- Achilles (VISA-A) selected as the primary clinical outcome measure.ResultsThirty participants were recruited, reflecting a 75% eligibility rate and 100% of eligible participants enrolled in the study. The retention rate at week 12 was 80%. At week six the mean VISA-A score improved by 18.1 (95% CI 10.2–26.0) in the intervention group and 7.7 (95% CI 0.3–14.9) in the control group, and 75% and 33% of participants in the intervention and control group respectively exceeded the minimal clinically important difference (MCID). At week 12 the mean VISA-A score from baseline improved by 22.25 (95% CI 12.52–31.98) in the intervention group and 16.5-(95% CI 8.47–24.53) in the control group, equating to 75% and 58% in each group respectively exceeding the MCID.ConclusionThe positive feasibility outcomes and exploratory data from the clinical outcome measures suggest that a larger scaled RCT is warranted to further investigate the impact of AOT in the rehabilitation of mid-portion AT.Trial registration ISRCTN58161116, first registered on the 23/12/2020.

Brain2Pix: Fully convolutional naturalistic video frame reconstruction from brain activity.

Reconstructing complex and dynamic visual perception from brain activity remains a major challenge in machine learning applications to neuroscience. Here, we present a new method for reconstructing naturalistic images and videos from very large single-participant functional magnetic resonance imaging data that leverages the recent success of image-to-image transformation networks. This is achieved by exploiting spatial information obtained from retinotopic mappings across the visual system. More specifically, we first determine what position each voxel in a particular region of interest would represent in the visual field based on its corresponding receptive field location. Then, the 2D image representation of the brain activity on the visual field is passed to a fully convolutional image-to-image network trained to recover the original stimuli using VGG feature loss with an adversarial regularizer. In our experiments, we show that our method offers a significant improvement over existing video reconstruction techniques.

The impact of nature video exposure on pro-environmental behavior: An experimental investigation.

We analyze whether exposure to a nature documentary increases pro-environmental behavior (PEB). We test this causal link in an experiment where subjects viewed a video featuring either an urban (control treatment) or a nature setting (nature treatment). We consider two types of behavior: a monetary donation to an environmental non-governmental organization (ENGO) that we call an eco-donation, and subsequently, a non-monetary decision (i.e., recycle or not recycle headphone protectors) that we call an eco-action. We find that virtual exposure to nature boosts both eco-donation and eco-action. Interestingly, the increase in PEB only occurs for individuals who express low environmental values. We did not find any negative or positive spillover effects on the eco-action. We finally provide robustness checks and discuss policy implications.

INTEGRATIVE SEMANTIC MAP AS AN ILLUSTRATIVE AND SCHEMATIC BASIS FOR THE DEVELOPMENT OF A PROFESSIONALLY SIGNIFICANT ENGLISH-LANGUAGE THESAURUS OF THE THEMATIC FIELD “LAW”

The authors of the article consider integrative semantic mapping as an illustrative and schematic basis for the law students’ development of a professionally oriented English-language thesaurus of the topical field “Law”. The aim of the research is to study further development of professional and communicative English-language competence which allows ensuring the implementation of pragmatic and professional communicative tasks in simulated situations of entering into legal relationship by the students.
 Methodology. The methodological framework for the research is the cognitive and generalizing methods, such as the theoretical analysis of research works in linguodidactics, pedagogy, cognitive linguistics, and the prognostic methods. The research material was an educational and methodological complex on the discipline “Legal English”, used as part of the Bachelor’s degree program 40.03.01 “Jurisprudence”.
 Results. The paper analyzes an algorithm for the development of abilities and skills of linguo-cognitive processing of professionally significative highly legal English texts of a static nature (for intermediate and upper-intermediate students), and English-language videos of a dynamic nature of the genre “courtroom drama” (for advanced and proficient students) using integrative semantic maps. Analyzing their application features allows us to emphasize their main function, which consists in optimizing the development of information about the law enforcement and legal proceedings in English-speaking countries, necessary for the effective intercultural communication of future practitioners in various branches of law.
 Conclusion. Forming and developing the English-language thesaurus of a professional lawyer is one of important areas of modern linguodidactics. It has been proved that the optimization of this process has become possible thanks to the use of new learning technologies, in particular, integrative semantic mapping as a tool for verbalization of acquired knowledge, which allows consciously applying cognitively mastered information, competently verbalizing it in intercultural professional communication, preventing and eliminating possible communication failures.

No-Reference Video Quality Assessment using novel hybrid features and two-stage hybrid regression for score level fusion

This paper presents a novel No-Reference Video Quality Assessment (NR-VQA) model that utilizes proposed 3D steerable wavelet transform-based Natural Video Statistics (NVS) features as well as human perceptual features. Additionally, we proposed a novel two-stage regression scheme that significantly improves the overall performance of quality estimation. In the first stage, transform-based NVS and human perceptual features are separately passed through the proposed hybrid regression scheme: Support Vector Regression (SVR) followed by Polynomial curve fitting. The two visual quality scores predicted from the first stage are then used as features for the similar second stage. This predicts the final quality scores of distorted videos by achieving score level fusion. Extensive experiments were conducted using five authentic and four synthetic distortion databases. Experimental results demonstrate that the proposed method outperforms other published state-of-the-art benchmark methods on synthetic distortion databases and is among the top performers on authentic distortion databases. The source code is available at https://github.com/anishVNIT/two-stage-vqa.

Virtual reality-based sensorimotor adaptation shapes subsequent spontaneous and naturalistic stimulus-driven brain activity.

Our everyday life summons numerous novel sensorimotor experiences, to which our brain needs to adapt in order to function properly. However, tracking plasticity of naturalistic behavior and associated brain modulations is challenging. Here, we tackled this question implementing a prism adaptation-like training in virtual reality (VRPA) in combination with functional neuroimaging. Three groups of healthy participants (N= 45) underwent VRPA (with a shift either to the left/right side, or with no shift), and performed functional magnetic resonance imaging (fMRI) sessions before and after training. To capture modulations in free-flowing, task-free brain activity, the fMRI sessions included resting-state and free-viewing of naturalistic videos. We found significant decreases in spontaneous functional connectivity between attentional and default mode (DMN)/fronto-parietal networks, only for the adaptation groups, more pronouncedly in the hemisphere contralateral to the induced shift. In addition, VRPA was found to bias visual responses to naturalistic videos: Following rightward adaptation, we found upregulation of visual response in an area in the parieto-occipital sulcus (POS) only in the right hemisphere. Notably, the extent of POS upregulation correlated with the size of the VRPA-induced after-effect measured in behavioral tests. This study demonstrates that a brief VRPA exposure can change large-scale cortical connectivity and correspondingly bias visual responses to naturalistic sensory inputs.

Spectral brain signatures of aesthetic natural perception in the α and β frequency bands.

During our everyday lives, visual beauty is often conveyed by sustained and dynamic visual stimulation, such as when we walk through an enchanting forest or watch our pets playing. Here, I devised an MEG experiment that mimics such situations: participants viewed 8 s videos of everyday situations and rated their beauty. Using multivariate analysis, I linked aesthetic ratings to 1) sustained MEG broadband responses and 2) spectral MEG responses in the α and β frequency bands. These effects were not accounted for by a set of high- and low-level visual descriptors of the videos, suggesting that they are genuinely related to aesthetic perception. My findings provide the first characterization of spectral brain signatures linked to aesthetic experiences in the real world.NEW & NOTEWORTHY In the real world, aesthetic experiences arise from complex and dynamic inputs. This study shows that such aesthetic experiences are represented in a spectral neural code: cortical α and β activity track our judgments of the aesthetic appearance of natural videos, providing a new starting point for studying the neural correlates of beauty through rhythmic brain activity.

S514 Virtual Reality Improves Symptoms of Functional Dyspepsia: Results of a Randomized, Controlled, Double-Blind Pilot Study

Introduction: Functional dyspepsia (FD) is a common functional GI disorder, characterized by symptoms of abdominal pain, nausea, and/or early satiety. Virtual reality (VR) is a promising intervention which has been shown to alleviate pain in various clinical settings. There are currently no studies assessing VR for the treatment of FD. The aim of this study was to investigate the effectiveness and safety of VR for the treatment of FD. Methods: This is a prospective, single-center, randomized, controlled, double-blinded study of adult patients with FD (Rome IV criteria). Enrolled patients were randomized 2:1 (experimental to control). Experimental patients were given a VR headset with software consisting of immersive audiovisual programs. Control patients were given an identical headset with 2-D nature videos. Patients were asked to use their headset at least daily and completed Patient Assessment of Gastrointestinal Disorders-Symptom Severity Index (PAGI-SYM) and Nepean Dyspepsia Index (NDI) questionnaires at their initial visit, after 1 week of use, and at the conclusion of the 2-week study. Paired sample t-tests were used to compare questionnaire scores. Results: Thirty-seven patients were enrolled in the study (27 experimental, 10 controls); demographic data are shown in Table (Table). Patients used the VR headset an average of 1.3 times/day for a mean of 23.2 minutes/day. Total PAGI-SYM scores significantly decreased for all patients [2.51 (baseline) to 0.62 (week 2); p=0.000]. Notably, patients in the experimental group had greater improvement in mean total PAGI-SYM scores [2.51 baseline) to 1.83 (week 2); mean difference -0.68; p=0.000], compared to the control group [2.50 (baseline) to 2.04 (week 2); mean difference -0.46; p=0.046]. Further, quality of life (QoL) significantly improved for all patients, as the total NDI QoL score increased from 40.97 (baseline) to 57.14 (week 2), p=0.000, though experimental patients saw greater improvement in QoL, compared to control patients (Figure). Seventeen patients (45.9%; 11 experimental, 6 controls) reported non-serious adverse effects (AE), with headache and dizziness being the most common; 1 experimental patient withdrew due to AEs (migraines). Conclusion: In the first randomized, controlled, double-blind study to assess VR for the treatment of FD, VR was shown to be safe and result in statistically significant overall improvement in symptoms and QoL in FD patients. We propose VR as a novel treatment for FD worthy of further investigation.Figure 1.: Total PAGI-SYM and NDI Quality of Life Scores Table 1. - Study Population Demographic Data Mean Age 45 years (± 14) Mean BMI 24.56 (± 6.79) SexFemaleMale 30 (81.1%)7 FD subtypeEPSPDSMixed EPS/PDS 12 (32.4%)20 (54.1%)5 (13.5%) Anxiety 17 (45.9%) Depression 9 (24.3%) PTSD 2 (5.4%) Fibromyalgia 4 (10.8%) Prior H. pylori infection 5 (13.5%) MedicationsPPIAntidepressantsGabapentinOpioidsNSAIDs 16 (43.2%)11 (29.7%)4 (10.8%)2 (5.4%)0 Alcohol Use 7 (18.9%) Marijuana Use 4 (10.8%)

Effects of Meaningful Action Observation Therapy on Occupational Performance, Upper Limb Function, and Corticospinal Excitability Poststroke: A Double-Blind Randomized Control Trial.

Introduction Action observation therapy (AOT) is a mirror neuron-based approach that has been recently used in poststroke rehabilitation. The main goal of this study was to investigate the effectiveness of AOT of occupations and tasks that are meaningful for chronic stroke patients on occupational performance, upper-extremity function, and corticospinal changes. Method A randomized control trial was designed to compare between experimental (n = 13) and control groups (n = 14). In both groups, the execution of meaningful tasks was practiced, but the videos of those tasks were just shown to the experiment group. Instead, patients in the control group watched nature videos as a placebo. Clinical outcomes were evaluated using the Canadian Occupational Performance Measure (COPM), Fugl-Meyer Assessment (FMA), Action Research Arm Test (ARAT), and Box-Block Test (BBT) on 3 occasions: baseline, post (at 4 weeks), and follow-up (at 8 weeks). The assessments of central motor conduction time (CMCT) for abductor policis brevis (APB) and extensor indicis (EI) were only recorded at baseline and posttreatment. Both assessors of clinical and neurophysiological outcomes were blinded to the allocation of subjects. Result Finally, the results of outcomes in 24 patients who completed the study were analyzed. In both groups, significant improvements after treatment were seen for most outcomes (p ≤ 0.05). These changes were persistent until follow-up. There were significant differences in COPM performance (p = 0.03) and satisfaction (p = 0.001) between the experimental and control groups. In contrast, other clinical assessments such as FMA, ARAT, and BBT did not show significant differences between the two treatments (p ≥ 0.05). The results of CMCT related to APB showed a more significant change in the experiment group compared to the control group (p = 0.022). There was no difference in change detected between the two groups for CMCT related to EI after treatments. Conclusion Observation and execution of meaningful activities can enhance the effects of simply practicing those activities on occupational performance/satisfaction and corticospinal excitability poststroke.

Through a Dog's Eyes: fMRI Decoding of Naturalistic Videos from the Dog Cortex.

Recent advancements using machine learning and functional magnetic resonance imaging (fMRI) to decode visual stimuli from the human and nonhuman cortex have resulted in new insights into the nature of perception. However, this approach has yet to be applied substantially to animals other than primates, raising questions about the nature of such representations across the animal kingdom. Here, we used awake fMRI in two domestic dogs and two humans, obtained while each watched specially created dog-appropriate naturalistic videos. We then trained a neural net (Ivis) to classify the video content from a total of 90 min of recorded brain activity from each. We tested both an object-based classifier, attempting to discriminate categories such as dog, human, and car, and an action-based classifier, attempting to discriminate categories such as eating, sniffing, and talking. Compared to the two human subjects, for whom both types of classifier performed well above chance, only action-based classifiers were successful in decoding video content from the dogs. These results demonstrate the first known application of machine learning to decode naturalistic videos from the brain of a carnivore and suggest that the dog's-eye view of the world may be quite different from our own.

Video Super-Resolution for Wireless Capsule Endoscopy Imaging Sensor

As a preferred diagnostic imaging sensor for small bowel diseases, wireless capsule endoscopy (WCE) suffers from poor video quality, which can be mitigated via video super-resolution (VSR) techniques. However, the promise of applying VSR into WCE applications has been offset by two significant challenges. Firstly, high-resolution (HR) counterparts of real endoscopic frames are unavailable for supervised learning. Secondly, utilizing temporal correlations to enhance recovery performance becomes difficult due to the poor video quality. To tackle these two challenges, we develop a novel training dataset and an effective alignment network for VSR on practical WCE applications. Specifically, clean natural videos and synthetic endoscopic videos are collected as the ground truth of the training dataset. Then, a set of complex degradation models is designed to generate their low-resolution (LR) counterparts for supervised learning. Although real HR endoscopic videos remain unknown during training, this complicated dataset enables the trained network to have a strong generalization capacity so that it can make reasonable inferences on real endoscopic videos. To utilize temporal correlations from WCE videos, a block-based temporal attention alignment network (BTAAN) is proposed to build accurate and robust correspondences among consecutive frames under poor video quality. This network exploits the global information from neighboring frames to capture large-scale motions, and suppresses the effect of unreliable points by a temporal attention mechanism. Moreover, its block-wise characteristic reduces the computational complexity significantly, which makes it hardware-friendly. Extensive experiments demonstrate the superiority of the proposed method over other state-of-the-art approaches on practical WCE applications.

Nature videos for PTSD: protocol for a mixed-methods feasibility study

ABSTRACT Background: Given the prevalence of post-traumatic stress disorder (PTSD), particularly among military personnel, new treatment approaches are needed. One may be virtual relaxation interventions, especially 360-degree nature videos, since studies have demonstrated their relaxation effects for healthy participants. If these relaxation effects can be reproduced in patients with PTSD, they may offer a viable tool to reduce distress and hyperarousal. Objective: This research protocol describes a planned study that will examine the relaxation effects of 360-degree nature videos for patients with PTSD. It will also investigate whether these relaxation effects differ depending on the hardware immersion level (head-mounted display [HMD] vs. PC screen) in comparison to a control condition in which patients only listen to natural sounds and do not view a video. Finally, the effect of each intervention’s dose duration (five vs. ten minutes) will be explored. Method: A counterbalanced, randomised, controlled, within-subject experiment will be conducted (sample size N = 36). Only soldiers aged 18 years or older with a primary diagnosis of PTSD will be included. Those with psychosis, substance dependence, a change in psychiatric medication within the last month, suicidal intent, and motion sickness will be excluded. All patients will experience the HMD, PC, and control conditions once for five or ten minutes. Self-reported relaxation measures will be collected before and after, and patients’ skin conductance level, heart rate, and heart rate variability will be assessed during each condition. Semi-structured interviews will be conducted to examine the patients’ experiences in detail. Conclusions: This feasibility study will provide initial evidence of whether viewing 360-degree nature videos via HMD or PC screen is relaxing for patients with PTSD and whether the effects are greater compared with the control condition. The study will also validate the dose duration and thereby informing a subsequent confirmatory interventional trial. Trial registration: DRKS00020277. HIGHLIGHTS This randomised controlled feasibility study will examine whether 360-degree nature videos are a suitable relaxation intervention for military personnel with post-traumatic stress disorder.

Impaired socio-affective, but intact socio-cognitive skills in patients with treatment-resistant, recurrent depression

BackgroundSocial withdrawal is a key symptom of depression. The resulting loss of social reinforcement in turn contributes to chronic, recurrent courses of the disease. However, it is not clear whether depressed patients have less motivation to socially interact, or whether their skills in doing so are impaired. The current study investigates potential skill deficits in patients with treatment-resistant depression (TRD). Methods15 TRD patients and 19 age- and sex-matched healthy controls performed the EmpaToM, a paradigm which includes naturalistic video stimuli of either neutral or emotional valence and which differentiates between socio-affective (affective empathy, compassion) and socio-cognitive (theory of mind) skills. ResultsControlling for the baseline affective state in neutral situations, TRD patients displayed significantly reduced affective empathy towards emotional situations compared to healthy controls. Furthermore, TRD patients were less compassionate in both neutral and emotional situations. In contrast, socio-cognitive skill performances did not differ between patients and healthy controls. LimitationsFurther studies might explore socio-affective and socio-cognitive skills in TRD patients using socio-affective/-cognitive tasks involving face-to-face social interactions. ConclusionOur study revealed a specific socio-affective deficit in TRD patients, while showing intact socio-cognitive skills. Patients were less able to affectively resonate with others (affective empathy) and exhibited generally reduced feelings of compassion. These deficits might interfere with providing and receiving social support. Our study contributes to a better understanding of the underlying causes of social withdrawal and stresses the need to specifically address pervasive socio-affective deficits in psychotherapy of TRD patients.

Mesh Denoising Based on Recurrent Neural Networks

Mesh denoising is a classical task in mesh processing. Many state-of-the-art methods are still unable to quickly and robustly denoise multifarious noisy 3D meshes, especially in the case of high noise. Recently, neural network-based models have played a leading role in natural language, audio, image, video, and 3D model processing. Inspired by these works, we propose a data-driven mesh denoising method based on recurrent neural networks, which learns the relationship between the feature descriptors and the ground-truth normals. The recurrent neural network has a feedback loop before entering the output layer. By means of the self-feedback of neurons, the output of a recurrent neural network is related not only to the current input but also to the output of the previous moments. To deal with meshes with various geometric features, we use k-means to cluster the faces of the mesh according to geometric similarity and train neural networks for each category individually in the offline learning stage. Each network model, acting similar to a normal regression function, will map the geometric feature descriptor of each facet extracted from the mesh to the denoised facet normal. Then, the denoised normals are used to calculate the new feature descriptors, which become the input of the next similar regression model. In this system, three normal regression modules are cascaded to generate the last facet normals. Lastly, the model’s vertex positions are updated according to the denoised normals. A large number of visual and numerical results have demonstrated that the proposed model outperforms the state-of-the-art methods in most cases.

Rules and exceptions: A Tolerance Principle account of the possessive suffix in Northern East Cree.

Debate around inflectional morphology in language acquisition has contrasted various rule- versus analogy-based approaches. This paper tests the rule-based Tolerance Principle (TP) against a new type of pattern in the acquisition of the possessive suffix -im in Northern East Cree. When possessed, each noun type either requires or disallows the suffix, which has a complex distribution throughout the lexicon. Using naturalistic video data from one adult and two children - Ani (2;01-4;03) and Daisy (3;08-5;10) - this paper presents two studies. Study 1 applies the TP to the input to extrapolate two possible sets of nested rules for -im and make predictions for child speech. Study 2 tests these predictions and finds that each child's production of possessives over time is largely consistent with the predictions of the TP. This paper finds the TP can account for the acquisition of the possessive suffix and discusses implications for language science and Cree language communities.

Practical protection against video data leakage via universal adversarial head

While online video sharing becomes more popular, it also causes unconscious leakage of personal information in the video retrieval systems like deep hashing. A snoop can collect more users’ private information from the video database by querying similar videos. This paper focuses on bypassing the deep video hashing based retrieval to prevent information from being maliciously collected. We propose universal adversarial head (UAH), which crafts adversarial query videos by prepending the original videos with a sequence of adversarial frames to perturb the normal hash codes in the Hamming space. This adversarial head can be generated only with a few natural videos, and mislead the retrieval system to return irrelevant videos when it is applied to most query videos. Furthermore, to obey the principle of information protection, we expand the proposed method to a data-free paradigm to generate the UAH, without access to users’ original videos. Extensive experiments demonstrate the effectiveness of our method in misleading deep video hashing under both white-box and black-box settings.