Meaningful Segments Research Articles

Reading travel guidebooks: Readership typologies using eye-tracking technology

Abstract The increasing popularity and capabilities of eye-tracking technology applied in a tourism setting have allowed for a deeper understanding of tourist behaviours and perceptions than ever before. Used in the context of travel guidebooks, this exploratory study combines eye-tracking technology with traditional survey and interview data to build an informed understanding of the variability in tourism guidebook reading behaviours. After a structured qualitative analysis with 22 participants tasked with planning a trip to Ottawa, Canada, the findings reveal four broad types of readers: holistic content examiners, visual passives, supplemental scanners, and visual dominants. Each group's characteristics and patterns are compared and contrasted in regard to time to first fixation, visit count, and total fixation duration in each of the 68 pages within the guide, revealing nuanced descriptions of how destination marketers may better design their travel guides to accommodate these various reader types. Practical implications are provided, as well as important implications for marketing researchers aiming to better understand how to use eye-tracking technology to organise their participants into meaningful segments.

Wavelet‐Based River Segmentation Using Hydraulic Control‐Preserving Water Surface Elevation Profile Properties

AbstractRiver flow models that consider satellite observations of the water surface elevation (WSE) require meaningful segmentation into reaches. Segmentation methods involve a trade‐off between the spatial sampling and error characteristics of satellite observations, both of which impact the flow model realism/accuracy. This paper investigates the spatial properties of WSE profiles (slope and concavity) and their link with hydraulic controls (HCs) and leverages this knowledge to advance river segmentation. Fine‐scale synthetic cases were analyzed to characterize HC effects on the remotely observable WSE properties. Following this, a wavelet‐based segmentation method was developed and subsequently tested on real rivers, including test cases with measurements characteristic of the Surface Water and Ocean Topography satellite mission. The results show that the local extrema of water surface (WS) concavity are good candidates for defining reach bounds while consistently preserving the HC signatures and local flow nonuniformities (deviations from equilibrium) from fine to large spatial scales.

Towards genre-specific frameworks for video summarisation: A survey

Video summarisation is characterised as the process of extracting meaningful frames or segments from a video that best represents the content of the whole video. The proposed framework surveys and categorizes the existing video summarisation models in the recent research works on the basis of genre. The most important phase of video summarisation is the detection of key frames or segments in the video. The strategy for identifying key frames or segments vary for each genre. The various genre analysed are user generated videos, movies and documentary, sports, surveillance, egocentric and informational talk videos with a total of more than 25 varying parameters significant to the respective genre. Comprehensive evaluations of the results obtained from the models are also included based on quantitative and qualitative parameters. The framework will help the user in deciding the technology to be adopted for video summarisation in a particular domain. The framework also aids in deciding the type of summary suitable for each genre and the available datasets in each genre for experimental analysis.

Segmentifier: Interactive Refinement of Clickstream Data

AbstractClickstream data has the potential to provide insights into e‐commerce consumer behavior, but previous techniques fall short of handling the scale and complexity of real‐world datasets because they require relatively clean and small input. We present Segmentifier, a novel visual analytics interface that supports an iterative process of refining collections of action sequences into meaningful segments. We present task and data abstractions for clickstream data analysis, leading to a high‐level model built around an iterative view‐refine‐record loop with outcomes of conclude with an answer, export segment for further analysis in downstream tools, or abandon the question for a more fruitful analysis path. Segmentifier supports fast and fluid refinement of segments through tightly coupled visual encoding and interaction with a rich set of views that show evocative derived attributes for segments, sequences, and actions in addition to underlying raw sequences. These views support fast and fluid refinement of segments through filtering and partitioning attribute ranges. Interactive visual queries on custom action sequences are aggregated according to a three‐level hierarchy. Segmentifier features a detailed glyph‐based visual history of the automatically recorded analysis process showing the provenance of each segment as an analysis path of attribute constraints. We demonstrate the effectiveness of our approach through a usage scenario with real‐world data and a case study documenting the insights gained by a corporate e‐commerce analyst.

STRUCTURAL SEGMENTATION OF POINT CLOUDS WITH VARYING DENSITY BASED ON MULTI-SIZE SUPERVOXELS

Abstract. Ground objects can be regarded as a combination of structures of different geometries. Generally, the structural geometries can be grouped into linear, planar and scatter shapes. A good segmentation of objects into different structures can help to interpret the scanned scenes and provide essential clues for subsequent semantic interpretation. This is particularly true for the terrestrial static and mobile laser scanning data, where the geometric structures of objects are presented in detail due to the close scanning distances. In consideration of the large data volume and the large variation in point density of such point clouds, this paper presents a structural segmentation method of point clouds to efficiently decompose the ground objects into different structural components based on supervoxels of multiple sizes. First, supervoxels are generated with sizes adaptive to the point density with minimum occupied points and minimum size constraints. Then, the multi-size supervoxels are clustered into different components based on a structural labelling result obtained via Markov random field. Two datasets including terrestrial and mobile laser scanning point clouds were used to evaluate the performance of the proposed method. The results indicate that the proposed method can effectively and efficiently classify the point clouds into structurally meaningful segments with overall accuracies higher than 96%, even with largely varying point density.

On a segmentation of Coopextra customers in Thu Duc district

Customer segmentation is the process of grouping customers based on similar characteristics such as behavior, shopping habits…so that businesses can do marketing to each customer group effectively and appropriately. Customer segmentation helps businesses determine different strategies and different marketing approaches to different groups. Customer segmentation helps marketers better understand customers as well as provide goals, strategies and marketing methods for different target groups. This paper aims to examine the customer segmentation using clustering method in statistics and unsupervised machine learning. The algorithms used are K-means and Elbow which are famous algorithms that have been successfully applied in many areas such as marketing, biology, library, insurance, finance... The purpose of clustering is to find meaningful market segments. However, the adoption and adjustment of parameters in the algorithms so as to find significant customer segmentations remain a challenge at present. In this paper, we used data of customers of Thu Duc CoopExtra and found significant customer segmentations which can be useful for more effective marketing and customer care by the supermarket.

How the Human Brain Segments Continuous Experience.

The formation and subsequent recall of coherent episodes from our past (i.e., our episodic memories) relies on the ability to parse the continuous stream of experience in our daily lives into meaningful chunks or segments ([Zacks et al., 2010][1]). This segmentation is thought to determine how

Understanding High-Speed Rail Passengers in China: A Segmentation Approach

China has established the world’s largest high-speed rail (HSR) system, which has fundamentally changed the way people travel in the domestic market. As China aims to double its HSR capacity in the next few years, the HSR population will continue to grow, which calls for an in-depth understanding of HSR passengers. While HSR has been of academic interest for many years, existing research has not provided meaningful demographic segmentation in the HSR context. This paper collected empirical data from HSR passengers in Beijing and Shanghai, the largest HSR markets in China, and performed a cluster analysis based collectively on three demographic variables—age, income, and education, which led to the formation of four segments—High-Ed Youths, Mature Travelers, New Starters, and Elite Travelers. Significant differences were found in terms of the passenger demographics and travel experiences across the four segments, to support the validity of the clustering solutions. The multivariate analysis of variance (MANOVA) test further revealed cross-segment differences in terms of passenger evaluation of five HSR variables—reasonableness of price, reliability, food choices, employee service, and likelihood for recommendation, suggesting the possibility of predicting passenger perceptions and behaviors based on their cluster membership. The findings demonstrate that passenger segmentation based on multiple demographic variables can provide deeper insights into the HSR population. For HSR providers in China, an understanding of the characteristics of the four passenger segments can assist them in developing service and communication strategies to cater to the different passenger needs.

A Study on Image Segmentation Techniques

Image processing is a technique to transform an image into digital form and implement some operations on it; in order to acquire an improved image or to abstract some useful information from it. It is a kind of signal exemption in which input is image, like video frame or photograph and output may be image or characteristics related with that image. Segmentation partitions an image into separate regions comprising each pixel with similar attributes. To be significant and useful for image analysis and clarification, the regions should powerfully relate to depicted objects or features of interest. Meaningful segmentation is the first step from low-level image processing converting a grey scale or color image into one or more other images to high-level image depiction in terms of objects, features, and scenes. The achievement of image analysis depends on reliability of segmentation, but an exact partitioning of an image is mostly a very challenging problem.

Sensor-to-body calibration procedure for clinical motion analysis of lower limb using magnetic and inertial measurement units

Magnetic and Inertial measurement units (MIMUs) have become exceedingly popular for ambulatory human motion analysis during the past two decades. However, measuring anatomically meaningful segment and joint kinematics requires virtual alignment of the MIMU frame with the anatomical frame of its corresponding segment. Therefore, this paper presents a simple calibration procedure, based on MIMU readouts, to align the inertial frame of the MIMU with the anatomical frames, as recommended by ISB. The proposed calibration includes five seconds of quiet standing in a neutral posture followed by ten consecutive hip flexions/extensions. This procedure will independently calibrate MIMUs attached to the pelvis, thigh, shank, and foot. The accuracy and repeatability of the calibration procedure and the 3D joint angle estimation were validated against the gold standard motion capture system by an experimental study with ten able-bodied participants. The procedure showed high test-retest repeatability in aligning the MIMU frame with its corresponding anatomical frame, i.e., the helical angle between the MIMU and anatomical frames did not significantly differ between the test and retest sessions (except for thigh MIMU). Compared to previously introduced procedures, this procedure attained the highest inter-participant repeatability (inter-participant coefficient of variations of the helical angle: 20.5–42.2%). Further, the proposed calibration would reduce the offset errors of the 3D joint angle estimation (up to 12.8 degrees on average) compared to joint angle estimation without calibration (up to 26.3 degrees on average). The proposed calibration enables MIMU to measure clinically meaningful gait kinematics.

Dynamic [18F]FET-PET/MRI using standard MRI-based attenuation correction methods

AimTo assess if tumour grading based on dynamic [18F]FET positron emission tomography/magnetic resonance imaging (PET/MRI) studies is affected by different MRI-based attenuation correction (AC) methods.MethodsTwenty-four patients with suspected brain tumours underwent dynamic [18F]FET-PET/MRI examinations and subsequent low-dose computed tomography (CT) scans of the head. The dynamic PET data was reconstructed using the following AC methods: standard Dixon-based AC and ultra-short echo time MRI-based AC (MR-AC) and a model-based AC approach. All data were reconstructed also using CT-based AC (reference). For all lesions and reconstructions, time-activity curves (TACs) and time to peak (TTP) were extracted using different region-of-interest (ROI) and volume-of-interest (VOI) definitions. According to the most common evaluation approaches, TACs were categorised into two or three distinct curve patterns. Changes in TTP and TAC patterns compared to PET using CT-based AC were reported.ResultsIn the majority of cases, TAC patterns did not change. However, TAC pattern changes as well as changes in TTP were observed in up to 8% and 17% of the cases when using different MR-AC methods and ROI/VOI definitions, respectively. However, these changes in TTP and TAC pattern were attributed to different delineations of the ROIs/VOIs in PET corrected with different AC methods.ConclusionPET/MRI using different MR-AC methods can be used for the assessment of TAC patterns in dynamic [18F]FET studies, as long as a meaningful delineation of the area of interest within the tumour is ensured.Key Points• PET/MRI using different MR-AC methods can be used for dynamic [18F]FET studies.• A meaningful segmentation of the area of interest needs to be ensured, mandating a visual validation of the delineation by an experienced reader.

Skeleton-Based Action Recognition With Key-Segment Descriptor and Temporal Step Matrix Model

Human action recognition based on skeleton has played a key role in various computer vision-related applications, such as smart surveillance, human-computer interaction, and medical rehabilitation. However, due to various viewing angles, diverse body sizes, and occasional noisy data, etc., this remains a challenging task. The existing deep learning-based methods require long time to train the models and may fail to provide an interpretable descriptor to code the temporal-spatial feature of the skeleton sequence. In this paper, a key-segment descriptor and a temporal step matrix model are proposed to semantically present the temporal-spatial skeleton data. First, a skeleton normalization is developed to make the skeleton sequence robust to the absolute body size and initial body orientation. Second, the normalized skeleton data is divided into skeleton segments, which are treated as the action units, combining 3D skeleton pose and the motion. Each skeleton sequence is coded as a meaningful and characteristic key segment sequence based on the key segment dictionary formed by the segments from all the training samples. Third, the temporal structure of the key segment sequence is coded into a step matrix by the proposed temporal step matrix model, and the multiscale temporal information is stored in step matrices with various steps. Experimental results on three challenging datasets demonstrate that the proposed method outperforms all the hand-crafted methods and it is comparable to recent deep learning-based methods.

Combination of Active Transfer Learning and Natural Language Processing to Improve Liver Volumetry Using Surrogate Metrics with Deep Learning.

To determine if weakly supervised learning with surrogate metrics and active transfer learning can hasten clinical deployment of deep learning models. By leveraging Liver Tumor Segmentation (LiTS) challenge 2017 public data (n = 131 studies), natural language processing of reports, and an active learning method, a model was trained to segment livers on 239 retrospectively collected portal venous phase abdominal CT studies obtained between January 1, 2014, and December 31, 2016. Absolute volume differences between predicted and originally reported liver volumes were used to guide active learning and assess accuracy. Overall survival based on liver volumes predicted by this model (n = 34 patients) versus radiology reports and Model for End-Stage Liver Disease with sodium (MELD-Na) scores was assessed. Differences in absolute liver volume were compared by using the paired Student t test, Bland-Altman analysis, and intraclass correlation; survival analysis was performed with the Kaplan-Meier method and a Mantel-Cox test. Data from patients with poor liver volume prediction (n = 10) with a model trained only with publicly available data were incorporated into an active learning method that trained a new model (LiTS data plus over- and underestimated active learning cases [LiTS-OU]) that performed significantly better on a held-out institutional test set (absolute volume difference of 231 vs 176 mL, P = .0005). In overall survival analysis, predicted liver volumes using the best active learning-trained model (LiTS-OU) were at least comparable with liver volumes extracted from radiology reports and MELD-Na scores in predicting survival. Active transfer learning using surrogate metrics facilitated deployment of deep learning models for clinically meaningful liver segmentation at a major liver transplant center.© RSNA, 2019Supplemental material is available for this article.

Semantic Labeling and Instance Segmentation of 3D Point Clouds Using Patch Context Analysis and Multiscale Processing.

We present a novel algorithm for semantic segmentation and labeling of 3D point clouds of indoor scenes, where objects in point clouds can have significant variations and complex configurations. Effective segmentation methods decomposing point clouds into semantically meaningful pieces are highly desirable for object recognition, scene understanding, scene modeling, etc. However, existing segmentation methods based on low-level geometry tend to either under-segment or over-segment point clouds. Our method takes a fundamentally different approach, where semantic segmentation is achieved along with labeling. To cope with substantial shape variation for objects in the same category, we first segment point clouds into surface patches and use unsupervised clustering to group patches in the training set into clusters, providing an intermediate representation for effectively learning patch relationships. During testing, we propose a novel patch segmentation and classification framework with multiscale processing, where the local segmentation level is automatically determined by exploiting the learned cluster based contextual information. Our method thus produces robust patch segmentation and semantic labeling results, avoiding parameter sensitivity. We further learn object-cluster relationships from the training set, and produce semantically meaningful object level segmentation. Our method outperforms state-of-the-art methods on several representative point cloud datasets, including S3DIS, SceneNN, Cornell RGB-D and ETH.

A semantic-based video scene segmentation using a deep neural network

Video scene segmentation is very important research in the field of computer vision, because it helps in efficient storage, indexing and retrieval of videos. Achieving this kind of scene segmentation cannot be done by just calculating the similarity of low-level features presented in the video; high-level features should also be considered to achieve a better performance. Even though much research has been conducted on video scene segmentation, most of these studies failed to semantically segment a video into scenes. Thus, in this study, we propose a Deep-learning Semantic-based Scene-segmentation model (called DeepSSS) that considers image captioning to segment a video into scenes semantically. First, the DeepSSS performs shot boundary detection by comparing colour histograms and then employs maximum-entropy-applied keyframe extraction. Second, for semantic analysis, using image captioning that benefits from deep learning generates a semantic text description of the keyframes. Finally, by comparing and analysing the generated texts, it assembles the keyframes into a scene grouped under a semantic narrative. That said, DeepSSS considers both low- and high-level features of videos to achieve a more meaningful scene segmentation. By applying DeepSSS to data sets from MS COCO for caption generation and evaluating its semantic scene-segmentation task results with the data sets from TRECVid 2016, we demonstrate quantitatively that DeepSSS outperforms other existing scene-segmentation methods using shot boundary detection and keyframes. What’s more, the experiments were done by comparing scenes segmented by humans and scene segmented by the DeepSSS. The results verified that the DeepSSS’ segmentation resembled that of humans. This is a new kind of result that was enabled by semantic analysis, which was impossible by just using low-level features of videos.

Semantic segmentation of trajectories with improved agent models for pedestrian behavior analysis

ABSTRACTIn this paper, we propose a method for semantic segmentation of pedestrian trajectories based on pedestrian behavior models, or agents. The agents model the dynamics of pedestrian movements in two-dimensional space using a linear dynamics model and common start and goal locations of trajectories. First, agent models are estimated from the trajectories obtained from image sequences. Our method is built on top of the Mixture model of Dynamic pedestrian Agents (MDA); however, the MDA's trajectory modeling and estimation are improved. Then, the trajectories are divided into semantically meaningful segments. The subsegments of a trajectory are modeled by applying a hidden Markov model using the estimated agent models. Experimental results with a real trajectory dataset show the effectiveness of the proposed method as compared to the well-known classical Ramer-Douglas-Peucker algorithm and also to the original MDA model.

Unsupervised Hierarchical Clustering Approach for Tourism Market Segmentation Based on Crowdsourced Mobile Phone Data.

Understanding tourism related behavior and traveling patterns is an essential element of transportation system planning and tourism management at tourism destinations. Traditionally, tourism market segmentation is conducted to recognize tourist’s profiles for which personalized services can be provided. Today, the availability of wearable sensors, such as smartphones, holds the potential to tackle data collection problems of paper-based surveys and deliver relevant mobility data in a timely and cost-effective way. In this paper, we develop and implement a hierarchical clustering approach for smartphone geo-localized data to detect meaningful tourism related market segments. For these segments, we provide detailed insights into their characteristics and related mobility behavior. The applicability of the proposed approach is demonstrated on a use case in the Province of Zeeland in the Netherlands. We collected data from 1505 users during five months using the Zeeland app. The proposed approach resulted in two major clusters and four sub-clusters which we were able to interpret based on their spatio-temporal patterns and the recurrence of their visiting patterns to the region.

Perceptions and Acceptance of Levonorgestrel Implants among Family Planning Clients within greater Banjul Area, The Gambia

A better understanding of women’s perception of levonorgestrel implants has the potential to help women and their partners freely choose the number and spacing of their children and thus, increase contraceptive prevalence. The contraceptive prevalence rate is very low in the Gambia; only 8 per cent of the women of reproductive age use modern methods of family planning and 1 per cent using traditional methods and an implant utilization rate of 3.9%. Therefore, the purpose of the study is centered on exploring the perceptions and acceptance level of Levonorgestrel Implants among family planning (FP) clients in Western Health Region One (WHR, 1). The perception and acceptance of implants among users is not well documented in the Gambia despite established knowledge of availability at little or no cost in most health facilities. A phenomenological qualitative study design was employed and twelve women were purposively selected from two hospitals to serve as participants. Focus group discussions were used as data collection method among six women per discussion. Qualitative data analysis method used for the study was the Editing Analysis Style where, the data were interpreted in search of meaningful segments and units. The data were analyzed using themes and categorizing the responses of the participants under each theme with verbatim quotes used wherever necessary. Sociodemographic findings revealed that majority of the women were between the ages of 30-39 (SD± 5.35) ranging from 25 to 45 years with a mean age of 32.75 years. More than half of the study participants had between 1-4 children (n = 7, 58.33), with a mean parity of 3.75 (SD±2.48). Almost all of them were Muslims (n =11, 91.66 %). All of them received some form of education with the highest level being secondary (Conventional) (n = 6, 50%). Majority were housewives (n= 6, 50%), married (n = 11, 91.66%). The findings showed that six themes emerged from both Focus Group Discussions where the majority of the respondents gave similar answers. The themes that were reported on were; efficacy, service provider’s attitude, husband/spousal approval, undesired effects, myths and misconceptions, and satisfaction. In conclusion, the use of Implants for family planning depends on clients’ perceptions. Therefore, understanding client’s perceptions on implants would help to improve prevalence, thus reducing unintended pregnancies, induced abortions, short birth intervals, fertility rates and maternal mortality and morbidity in the country.

Learning structural motif representations for efficient protein structure search

MotivationGiven a protein of unknown function, fast identification of similar protein structures from the Protein Data Bank (PDB) is a critical step for inferring its biological function. Such structural neighbors can provide evolutionary insights into protein conformation, interfaces and binding sites that are not detectable from sequence similarity. However, the computational cost of performing pairwise structural alignment against all structures in PDB is prohibitively expensive. Alignment-free approaches have been introduced to enable fast but coarse comparisons by representing each protein as a vector of structure features or fingerprints and only computing similarity between vectors. As a notable example, FragBag represents each protein by a ‘bag of fragments’, which is a vector of frequencies of contiguous short backbone fragments from a predetermined library. Despite being efficient, the accuracy of FragBag is unsatisfactory because its backbone fragment library may not be optimally constructed and long-range interacting patterns are omitted.ResultsHere we present a new approach to learning effective structural motif presentations using deep learning. We develop DeepFold, a deep convolutional neural network model to extract structural motif features of a protein structure. We demonstrate that DeepFold substantially outperforms FragBag on protein structural search on a non-redundant protein structure database and a set of newly released structures. Remarkably, DeepFold not only extracts meaningful backbone segments but also finds important long-range interacting motifs for structural comparison. We expect that DeepFold will provide new insights into the evolution and hierarchical organization of protein structural motifs.Availability and implementation https://github.com/largelymfs/DeepFold

Characteristics and perceptions of twice-weekly webinars for primary care physicians in Japan: a qualitative study.

ObjectivesTo explore the characteristic features and perceived value of free twice-weekly webinars predominantly focusing on the continuing professional development of primary care physicians in Japan. MethodsIn this qualitative study, we conducted a focus group of the webinars’ participants (n=6, a purposive sample). The discussion was recorded, with the recording subsequently transcribed, separated into meaningful segments and then open-coded until thematic saturation was reached. Concepts were generated through selective coding. Finally, the extracted concepts were grouped into categories. ResultsExtracted concepts were grouped into five categories: technological breakthroughs, the creation of learning opportunities, external interaction, stimulation of internal interaction, and the advantages and disadvantages of nationwide expansion. The webinars were perceived to provide a comfortable learning climate, enabling physicians to teach one another, share their experiences and become virtual colleagues. The chat system stimulated real-time interaction between both a main speaker and participants and the participants. Participants were able to ask questions or give comments in a stress-free atmosphere. The webinars were found to elicit real-time, internal interaction within participating sites without interrupting the sessions. Participants also highly valued the absence of commercial sponsorship. The expansion of the webinars raised two concerns: the possibility of speakers becoming nervous and the increased burden on the organizers. ConclusionsThe webinars have successfully allowed sharing of unbiased information and experiences in a comfortable, multifaceted interactive learning environment, enabling participants to feel connected. The chat system permitted interaction not feasible in face-to-face learning opportunities and has shown great promise as a means of online medical education.