Perceptual Region Research Articles

Fine-grained image classification method based on hybrid attention module.

To efficiently capture feature information in tasks of fine-grained image classification, this study introduces a new network model for fine-grained image classification, which utilizes a hybrid attention approach. The model is built upon a hybrid attention module (MA), and with the assistance of the attention erasure module (EA), it can adaptively enhance the prominent areas in the image and capture more detailed image information. Specifically, for tasks involving fine-grained image classification, this study designs an attention module capable of applying the attention mechanism to both the channel and spatial dimensions. This highlights the important regions and key feature channels in the image, allowing for the extraction of distinct local features. Furthermore, this study presents an attention erasure module (EA) that can remove significant areas in the image based on the features identified; thus, shifting focus to additional feature details within the image and improving the diversity and completeness of the features. Moreover, this study enhances the pooling layer of ResNet50 to augment the perceptual region and the capability to extract features from the network's less deep layers. For the objective of fine-grained image classification, this study extracts a variety of features and merges them effectively to create the final feature representation. To assess the effectiveness of the proposed model, experiments were conducted on three publicly available fine-grained image classification datasets: Stanford Cars, FGVC-Aircraft, and CUB-200-2011. The method achieved classification accuracies of 92.8, 94.0, and 88.2% on these datasets, respectively. In comparison with existing approaches, the efficiency of this method has significantly improved, demonstrating higher accuracy and robustness.

Perception-Oriented U-Shaped Transformer Network for 360-Degree No-Reference Image Quality Assessment

Generally, 360-degree images have absolute senses of reality and three-dimensionality, providing a wide range of immersive interactions. Due to the novel rendering and display technology of 360-degree images, they have more complex perceptual characteristics than other images. It is challenging to perform comprehensive image quality assessment (IQA) learning by simply stacking multichannel neural network architectures for pre/postprocessing, compression, and rendering tasks. To thoroughly learn the global and local features in 360-degree images, reduce the complexity of multichannel neural network models and simplify the training process, this paper proposes a joint architecture with user perception and an efficient transformer dedicated to 360-degree no-reference (NR) IQA. The input of the proposed method is a 360-degree cube map projection (CMP) image. Furthermore, the proposed 360-degree NRIQA method includes a saliency map-based non-overlapping self-attention selection module and a U-shaped transformer (U-former)-based feature extraction module to account for perceptual region importance and projection distortion. The transformer-based architecture and the weighted average technique are jointly utilized for predicting local perceptual quality. Experimental results obtained on widely used databases show that the proposed model outperforms other state-of-the-art methods in NR 360-degree image quality evaluation cases. Furthermore, a cross-database evaluation and an ablation study also demonstrate the inherent robustness and generalization ability of the proposed model.

Reproduction of the identity of a region: perceptual regions based on formal and functional regions and their boundaries

ABSTRACT This paper introduces a way to study the reproduction of (the identity of) a region through the concept of perceptual region. Perceptual region—revised here in light of the institutionalization of regions theory and thus comprehended as the subjective image of a region in the mind of an individual person—connects certain ‘European’ and ‘American’ regional traditions. Investigating the institutions imprinted in perceptual regions exposes on what basis people construct regions and what (re)produces such regions. This case study deals with the perception of the Bohemian–Moravian boundary in the minds of 454 borderland inhabitants, surveyed with questionnaires. Among the revealed institutions of the former administrative regions of Bohemia and Moravia, there are various formal/homogeneous and functional regions as well as their boundaries—they all reproduce the examined regions. Besides demonstrating that regions are multiple realities (where both space and time matter), social constructs, and dynamic processes, the article discusses the interrelationship of the abovementioned region types and highlights perceptual regions as essential for the region’s existence. Furthermore, it suggests that focusing on ordinary people’s perceptions may develop knowledge not only about the concept of region but also the concepts of the regional identity of people, resistance identity, and regionalism.

Continuous Use Intention of Mobile Social Network Information Service Based on User Behavior Perception.

In order to improve the willingness of continuous use of mobile social network information services, this study combines user behavior perception to analyze the continuous use of mobile social network information services and proposes a data coverage optimization strategy based on service quality perception. Furthermore, this study measures participants' regional preferences based on the duration of participants in the perceptual region and the number of historical perceptual tasks completed on the perceptual region. In addition, this study designs a perceptual data coverage optimization algorithm to optimize the perceptual data coverage and ensure the real-time validity of the perceptual data. Through algorithm research and systematic evaluation, it can be seen that the continuous use willingness system of mobile social network information service based on user behavior perception can basically meet the actual needs.

Pasadena: Perceptually Aware and Stealthy Adversarial Denoise Attack

Image denoising can remove natural noise that widely exists in images captured by multimedia devices due to low-quality imaging sensors, unstable image transmission processes, or low light conditions. Recent works also find that image denoising benefits the high-level vision tasks, <i>e.g</i>., image classification. In this work, we try to challenge this common sense and explore a totally new problem, <i>i.e</i>., whether the image denoising can be given the capability of fooling the state-of-the-art deep neural networks (DNNs) while enhancing the image quality. To this end, we initiate the very first attempt to study this problem from the perspective of adversarial attack and propose the <i>adversarial denoise attack</i>. More specifically, our main contributions are three-fold: <i>First</i>, we identify a new task that stealthily embeds attacks inside the image denoising module widely deployed in multimedia devices as an image post-processing operation to simultaneously enhance the visual image quality and fool DNNs. <i>Second</i>, we formulate this new task as a kernel prediction problem for image filtering and propose the <i>adversarial-denoising kernel prediction</i> that can produce adversarial-noiseless kernels for effective denoising and adversarial attacking simultaneously. <i>Third</i>, we implement an adaptive <i>perceptual region localization</i> to identify semantic-related vulnerability regions with which the attack can be more effective while not doing too much harm to the denoising. We name the proposed method as <i>Pasadena</i> (Perceptually Aware and Stealthy Adversarial DENoise Attack) and validate our method on the NeurIPS&#x2019;17 adversarial competition dataset, CVPR2021-AIC-VI: unrestricted adversarial attacks on ImageNet, and Tiny-ImageNet-C dataset. The comprehensive evaluation and analysis demonstrate that our method not only realizes denoising but also achieves a significantly higher success rate and transferability over state-of-the-art attacks.

Prediction learning in adults with autism and its molecular correlates

BackgroundAccording to Bayesian hypotheses, individuals with Autism Spectrum Disorder (ASD) have difficulties making accurate predictions about their environment. In particular, the mechanisms by which they assign precision to predictions or sensory inputs would be suboptimal in ASD. These mechanisms are thought to be mostly mediated by glutamate and GABA. Here, we aimed to shed light on prediction learning in ASD and on its neurobiological correlates.MethodsTwenty-six neurotypical and 26 autistic adults participated in an associative learning task where they had to learn a probabilistic association between a tone and the rotation direction of two dots, in a volatile context. They also took part in magnetic resonance spectroscopy (MRS) measurements to quantify Glx (glutamate and glutamine), GABA + and glutathione in a low-level perceptual region (occipital cortex) and in a higher-level region involved in prediction learning (inferior frontal gyrus).ResultsNeurotypical and autistic adults had their percepts biased by their expectations, and this bias was smaller for individuals with a more atypical sensory sensitivity. Both groups were able to learn the association and to update their beliefs after a change in contingency. Interestingly, the percentage of correct predictions was correlated with the Glx/GABA + ratio in the occipital cortex (positive correlation) and in the right inferior frontal gyrus (negative correlation). In this region, MRS results also showed an increased concentration of Glx in the ASD group compared to the neurotypical group.LimitationsWe used a quite restrictive approach to select the MR spectra showing a good fit, which led to the exclusion of some MRS datasets and therefore to the reduction of the sample size for certain metabolites/regions.ConclusionsAutistic adults appeared to have intact abilities to make predictions in this task, in contrast with the Bayesian hypotheses of ASD. Yet, higher ratios of Glx/GABA + in a frontal region were associated with decreased predictive abilities, and ASD individuals tended to have more Glx in this region. This neurobiological difference might contribute to suboptimal predictive mechanisms in ASD in certain contexts.

Transformation of the identity of a region: Theory and the empirical case of the perceptual regions of Bohemia and Moravia, Czech Republic

Abstract By using the concept of perceptual region – an essential part of the identity of a region and a part of every person’s mental map – this paper demonstrates a way to examine the understudied transformation of (the identity of) a region and, specifically, its territorial shape (boundaries). This concept effectively fuses the “institutionalisation of regions” theory and the methodologies of behavioural geography. This case study of the perceptual regions of Bohemia and Moravia shows how and why these historical regions and their boundary/boundaries developed, after a significant deinstitutionalisation by splitting into smaller regions in an administrative reform. Many people now perceive the Bohemian-Moravian boundary according to the newly-emerged regional boundaries, which often ignore old (historical) boundaries. Thus, the territorial shape of Bohemia and Moravia is transformed, with the Vysočina Region emerging as one of the new regions to witness the most eroded consciousness concerning these historical regions and their boundaries. The impact of administrative reforms on the perception and thus also the transformation of regions and boundaries is obvious, but the results also suggest that the more radical the administrative changes (in terms of toponyms and boundary mismatches), the fuzzier the collective perceptions of historical boundaries become, as well as peoples’ consciousness of historical regions.

Neural representations of phonology in temporal cortex scaffold longitudinal reading gains in 5- to 7-year-old children

The objective of this study was to investigate whether phonological processes measured through brain activation are crucial for the development of reading skill (i.e. scaffolding hypothesis) and/or whether learning to read words fine-tunes phonology in the brain (i.e. refinement hypothesis). We specifically looked at how different grain sizes in two brain regions implicated in phonological processing played a role in this bidirectional relation. According to the dual-stream model of speech processing and previous empirical studies, the posterior superior temporal gyrus (STG) appears to be a perceptual region associated with phonological representations, whereas the dorsal inferior frontal gyrus (IFG) appears to be an articulatory region that accesses phonological representations in STG during more difficult tasks. 36 children completed a reading test outside the scanner and an auditory phonological task which included both small (i.e. onset) and large (i.e. rhyme) grain size conditions inside the scanner when they were 5.5–6.5 years old (Time 1) and once again approximately 1.5 years later (Time 2). To study the scaffolding hypothesis, a regression analysis was carried out by entering brain activation in either STG or IFG for either small (onset > perceptual) or large (rhyme > perceptual) grain size phonological processing at T1 as the predictors and reading skill at T2 as the dependent measure, with several covariates of no interest included. To study the refinement hypothesis, the regression analysis included reading skill at T1 as the predictor and brain activation in either STG or IFG for either small or large grain size phonological processing at T2 as the dependent measures, with several covariates of no interest included. We found that only posterior STG, regardless of grain size, was predictive of reading gains. Parallel models with only behavioral accuracy were not significant. Taken together, our results suggest that the representational quality of phonology in temporal cortex is crucial for reading development. Moreover, our study provides neural evidence supporting the scaffolding hypothesis, suggesting that brain measures of phonology could be helpful in early identification of reading difficulties.

Perceptual Effects of Nasal Cue Modification

Acoustic or perceptual cues used for speech perception can be very helpful in almost all areas of speech signal processing. A new methodology 3-Dimensional-Deep Search and a new visualized intelligible time-frequency computerbased model AI-gram have been developed and are being researched since the last several years (Human Speech Recognition (HSR) research group at the University of Illinois Urbana-Champaign) for isolation of stable perceptual cues of consonants. The perceptual cues of nasal consonants [1] have been successfully found considering these techniques [1]. The previous work is extended by assessing the changes in nasal sound perception and cue region is modified by using digital signal processing method. The amplitude of the perceptual cue region is amplified, attenuated or ignored completely and then the perception score is measured. A high correlation between the amplitude of the cue region and the perception score is found. The intelligibility of the entire token is increased or decreased approximately in a similar fashion as the cue region modified amplitude which is measured by the MMSE shift of the perceptual score curve. This validates that the regions identified are perceptual cue regions for nasal consonants. The digital signal processing method proposed can be used as a new approach for enhancing speech signal in noisy conditions.

표준 부호 디지트 코딩을 이용한 비가청 이진 랜덤 신호 발생에 관한 연구

디지털 워터마킹은 인간의 감각으로 감지할 수 없고 통계적인 방법에 의하여 검출되지 않는 정보를 디지털 데이터에 삽입하는 기술이다. 일반적으로 디지털 음향 신호에 대한 워터마킹은 의미 있는 정보를 갖는 이진 시퀀스를 원래의 디지털 음향 신호에 삽입하여 구현한다. 그러나 삽입된 이진 정보는 원래의 디지털 음향 신호의 관점에서는 잡음으로 작용하여 원래의 음향 신호를 왜곡 시키거나 열화 시키는 원인으로 작용한다. 본 논문에서는 인간의 가청 주파수 영역에서 이진 정보의 삽입으로 인한 원 음향 정보의 왜곡을 최소화시킬 수 있는 이진 정보 발생 방법을 제안하였다. 제안한 방법은 표준부호 디지트 코딩을 이용한 방법으로서 인접한 데이터간의 주파수 간섭을 최소화 하여 인간의 가청 주파수 영역에서의 이진 정보 시퀀스의 영향을 최소화 하도록 구현하였다. 제안한 방법은 일반적인 이진 정보발생 방법과 주파수 분석 비교를 통하여 성능을 평가하였다. Digital watermarking is imperceptible and statistically undetectable information embeds into digital data. Most information in digital audio watermarking schemes have used binary random sequences. The embedded binary random sequence distorts and modifies the original data while it plays a vital role in security. In this paper, a binary random sequence to improve imperceptibility in perceptual region of the human auditory system is proposed. The basic idea of this work is a modification of a binary random sequence according to the frequency analysis of adjacent binary digits that have different signs in the sequence. The canonical signed digit code (CSDC) is also applied to modify a general binary random sequence and the pair-matching function between original and its modified version. In our experiment, frequency characteristics of the proposed binary random sequence was evaluated and analyzed by Bark scale representation of frequency and frequency gains.

Towards perceptual video cropping with curve fitting

Coming with the proliferation of mobile devices, the issue of displaying videos with varying resolutions and aspect ratios has become emerging. To this end, state-of-the-art works typically resort to video retargeting or cropping, i.e. adaptively adjusting the display regions and contents within each video frame to fit for the mobile device screens. However, most existing approaches retain on evaluating the objective visual statistics to select salient parts to be cropped within each frame, with time-consuming constraints to ensure the temporal consistency, which is indeed unsuitable to cope with the massive videos uploaded daily from user-contributed platforms. In this paper, we propose an efficient yet robust video cropping algorithm which merits in two-fold: First, a perceptual Region of Interest selection approach is proposed by employing a novel Interesting Region Score, which is a linear combination of Rate of Focused Attention, Total Saliency Score, and Bias from Center Penalty. Second, we adopt fast curve fitting to seek for an optimal cropping sequence over a set of consecutive video frames within a shot, which avoids the complex optimization with temporal constraints among adjacent frames. Experiments on video sequences from Backkom cartoon and Iron Man II with comparisons against several alternative approaches demonstrate the efficiency and robustness of the proposed algorithm.

An effect of phoneme frequency on stop place perception by English-speaking and Japanese-speaking listeners.

The effects of phoneme frequency on stop place perception were examined. In English, [t] is more frequently observed than [k] while the opposite is true in Japanese. If the sound frequency affects the phoneme perception, English listeners would identify more of the ambiguous [t]-[k] stimuli as “t” than do Japanese listeners. In this study, a 4IAX discrimination experiment and a phoneme boundary decision experiment were conducted with English-speaking and Japanese-speaking listeners to test this hypothesis. The stimuli of the two experiments were created by the Klatt synthesizer. [ta], [ka], [to], and [ko] were recorded by a native Japanese speaker and a native English speaker, respectively. Then, [ta], [ka], [to], and [ko] were synthesized based on the parameter values obtained from the measurements of the recorded tokens. They were used as the end-point tokens of a nine-step continuum. The rest of the tokens were created by manipulating frequency parameters of the synthesizer. Four sets of [k]-[t] continua were created (two languages: English and Japanese, and two vowel conditions: [a] and [o]). A preliminary analysis of a subset of data demonstrated that the more frequent sound in each language occupied a larger perceptual region of the [a] vowel continua.

Peripheral perception mechanism of ultrasonic hearing

Ultrasound can be perceived by bone conduction, and its characteristics differ from those of air-conducted audible sound (ACAS) in some respects. Despite many studies on ultrasonic hearing, the details have not yet been clarified. In this study, to elucidate the perception mechanism, the masking of bone-conducted ultrasound (BCU) produced by ACAS and the sensitivity of BCU in hearing impaired subjects were evaluated. We found that BCU was masked by high frequency ACAS, especially in the frequency range of 10–14 kHz. The most effective masker frequency depended on masker intensity. For hearing impaired subjects, the pure tone thresholds at 1–8 kHz and the maximum audible frequencies at cut-off intensities of 70–100 dB HL were significantly associated with the BCU threshold (p < 0.01 or p < 0.05). No subjects with estimated total loss of the inner hair cell system in the cochlear basal turn could hear BCU. These results suggest the peripheral perceptual region to be located in the cochlea. The results of masking show the faster excitation spread to the lower frequency range, depending on the intensity. This faster excitation spread may be due to nonlinearity in cochlear mechanics, which may work even without cochlear amplifier, and induce unique characteristics of BCU.

Memory for motion and spatial location is mediated by contralateral and ipsilateral motion processing cortex

Memory and perception have been associated with common sensory cortical activity. However, previous studies have only investigated memory and perception effects associated with a single feature (i.e., spatial location or color). The aim of the present functional magnetic resonance imaging (fMRI) and transcranial magnetic stimulation (TMS) study was to assess whether memory for multiple (two) features would produce sensory cortical activity that mirrored perceptual processing of the same features. During encoding, moving or stationary abstract shapes were presented to the right or left of fixation. During retrieval, shapes were presented at fixation and participants classified each item as previously in motion or stationary within the right or left visual field. Memory for items in motion, regardless of spatial location, produced fMRI activity in perceptual motion processing region MT+. Memory for motion and spatial location produced contralateral and ipsilateral fMRI activity in perceptual motion processing sub-region MT. Following TMS to MT, memory for motion was impaired, but performance did not differ between the contralateral and ipsilateral visual fields. The present results are consistent with previous findings in that memory for motion produced fMRI activity in MT+ and was impaired following TMS to MT. However, memory for motion and spatial location produced contralateral and ipsilateral fMRI and TMS effects, deviating from the primarily contralateral perceptual processing organization of MT. The present evidence suggests that during memory for motion and spatial location only motion information is coded in motion processing cortex, while previous findings suggest spatial location information is coded in earlier extrastriate cortex.

The Sound of Concepts: Four Markers for a Link between Auditory and Conceptual Brain Systems

Traditionally, concepts are conceived as abstract mental entities distinct from perceptual or motor brain systems. However, recent results let assume modality-specific representations of concepts. The ultimate test for grounding concepts in perception requires the fulfillment of the following four markers: conceptual processing during (1) an implicit task should activate (2) a perceptual region (3) rapidly and (4) selectively. Here, we show using functional magnetic resonance imaging and recordings of event-related potentials, that acoustic conceptual features recruit auditory brain areas even when implicitly presented through visual words. Fulfilling the four markers, the findings of our study unequivocally link the auditory and conceptual brain systems: recognition of words denoting objects, for which acoustic features are highly relevant (e.g.,"telephone"), ignited cell assemblies in posterior superior and middle temporal gyri (pSTG/MTG) within 150 ms that were also activated by sound perception. Importantly, activity within a cluster of pSTG/MTG increased selectively as a function of acoustic, but not of visual and action-related feature relevance. The implicitness of the conceptual task, the selective modulation of left pSTG/MTG activity by acoustic feature relevance, the early onset of this activity at 150 ms and its anatomical overlap with perceptual sound processing are four markers for a modality-specific representation of auditory conceptual features in left pSTG/MTG. Our results therefore provide the first direct evidence for a link between perceptual and conceptual acoustic processing. They demonstrate that access to concepts involves a partial reinstatement of brain activity during the perception of objects.

Fast perceptual region tracking with coding-depth sensitive access for stream transcoding

Object-based bit allocation can result in significant improvement in the perceptual quality of extremely compressed video. However, real-time video object detection in large format high fidelity video is computationally daunting. Most algorithms begin with extensive use of classical bit analysis, and thus remain computationally heavy. Based on some recent results in human visual perception, in this paper, we present an experimental visual region tracking algorithm particularly designed for perceptual stream transcoding. This exploits the cue order observed in human visual perception to achieve very high computation speed as well as tracking efficiency. Rather than begin processing from pixel level or using any pixel level processing at all, it employs high level motion cue and block shape cue analysis to identify signatures of various relative movements between object of interest, scene background and the camera on the motion vector set, and from there it identifies objects. It then uses predictive filters to track the regions. The result is a fast yet highly effective perceptual region tracking algorithm that can operate in stream rate and track regions of perceptually significant object despite camera movements such as zoom, panning and translation. The technique is not specific to any special class of objects. We have implemented this algorithm in a live ISO-13818/MPEG-2 perceptual transcoder. In this paper, we share the performance of this implementation. This fast object-aware video rate transcoder is particularly suitable for live streaming and can convert a regular stream into a perceptually coded video stream.