Pub Date : 2019-03-11DOI: 10.1109/PERCOMW.2019.8730709
A. G. Zadeh, Puya Ghazizadeh, S. Olariu
Vehicular Clouds (VC) was inspired by the realization that the current vehicles are endowed with powerful sensing devices, storage, and computing resources and these resources are often underutilized. In this paper, we provide the reasoning for a communication protocol for vehicle-to-infrastructure (V2I) communications in Vehicular Cloud Computing systems. We first explain the structure of the proposed protocol in detail and then provide analytical predictions and simulation results to investigate the accuracy of our predictions.
{"title":"Reasoning About a Communication Protocol for Vehicular Cloud Computing Systems","authors":"A. G. Zadeh, Puya Ghazizadeh, S. Olariu","doi":"10.1109/PERCOMW.2019.8730709","DOIUrl":"https://doi.org/10.1109/PERCOMW.2019.8730709","url":null,"abstract":"Vehicular Clouds (VC) was inspired by the realization that the current vehicles are endowed with powerful sensing devices, storage, and computing resources and these resources are often underutilized. In this paper, we provide the reasoning for a communication protocol for vehicle-to-infrastructure (V2I) communications in Vehicular Cloud Computing systems. We first explain the structure of the proposed protocol in detail and then provide analytical predictions and simulation results to investigate the accuracy of our predictions.","PeriodicalId":437017,"journal":{"name":"2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124696047","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-03-11DOI: 10.1109/PERCOMW.2019.8730727
Aimilia-Myriam Michail, D. Gavalas
Food is far more than nourishment to humans; it can be a gastrimargic experience, an expressive form of art or even a social manifestation. Hence, applications that allow searching for best food options have become essential part of food experience for many. Technological advancements have brought the opportunity for custom search of food options at our fingertips. Tourist applications increasingly incorporate food search as an integral functional element, as gastronomy becomes an indispensable part of the travelling. However, most applications emphasize on the venue neglecting the menu options which is the primitive reason for food searching. The main objective of our work is to create an innovative food searching application tailored to gastronomic tourism that focuses on food rather than the venue. Our application adopts crowdsourcing principles, namely it relies on users to contribute content. Game elements are employed to motivate users in uploading accurate and qualitative food recommendations and sharing their food experiences in a social media-like fashion. The prototype implementation and the evaluation process provided us with valuable insights for the development of alike applications. Our findings are useful to application designers so as to effectively support gastronomic experiences and can be worthwhile for anyone planning to invest on gastronomic tourism or build a crowdsourcing platform.
{"title":"Bucketfood: A Crowdsourcing Platform for Promoting Gastronomic Tourism","authors":"Aimilia-Myriam Michail, D. Gavalas","doi":"10.1109/PERCOMW.2019.8730727","DOIUrl":"https://doi.org/10.1109/PERCOMW.2019.8730727","url":null,"abstract":"Food is far more than nourishment to humans; it can be a gastrimargic experience, an expressive form of art or even a social manifestation. Hence, applications that allow searching for best food options have become essential part of food experience for many. Technological advancements have brought the opportunity for custom search of food options at our fingertips. Tourist applications increasingly incorporate food search as an integral functional element, as gastronomy becomes an indispensable part of the travelling. However, most applications emphasize on the venue neglecting the menu options which is the primitive reason for food searching. The main objective of our work is to create an innovative food searching application tailored to gastronomic tourism that focuses on food rather than the venue. Our application adopts crowdsourcing principles, namely it relies on users to contribute content. Game elements are employed to motivate users in uploading accurate and qualitative food recommendations and sharing their food experiences in a social media-like fashion. The prototype implementation and the evaluation process provided us with valuable insights for the development of alike applications. Our findings are useful to application designers so as to effectively support gastronomic experiences and can be worthwhile for anyone planning to invest on gastronomic tourism or build a crowdsourcing platform.","PeriodicalId":437017,"journal":{"name":"2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)","volume":"93 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123129801","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-03-11DOI: 10.1109/PERCOMW.2019.8730792
Fernando Moya Rueda, S. Lüdtke, Max Schröder, Kristina Yordanova, T. Kirste, G. Fink
Activity recognition (AR) plays an important role in situation aware systems. Recently, deep learning approaches have shown promising results in the field of AR. However, their predictions are overconfident even in cases when the action class is incorrectly recognized. Moreover, these approaches provide information about an action class but not about the user context, such as location and manipulation of objects. To address these problems, we propose a hybrid AR architecture that combines deep learning with symbolic models to provide more realistic estimation of the classes and additional contextual information. We test the approach on a cooking dataset, describing the preparation of carrots soup. The results show that the proposed approach performs comparable to state of the art deep models inferring additional contextual properties about the current activity. The proposed approach is a first attempt to bridge the gap between deep learning and symbolic modeling for AR.
{"title":"Combining Symbolic Reasoning and Deep Learning for Human Activity Recognition","authors":"Fernando Moya Rueda, S. Lüdtke, Max Schröder, Kristina Yordanova, T. Kirste, G. Fink","doi":"10.1109/PERCOMW.2019.8730792","DOIUrl":"https://doi.org/10.1109/PERCOMW.2019.8730792","url":null,"abstract":"Activity recognition (AR) plays an important role in situation aware systems. Recently, deep learning approaches have shown promising results in the field of AR. However, their predictions are overconfident even in cases when the action class is incorrectly recognized. Moreover, these approaches provide information about an action class but not about the user context, such as location and manipulation of objects. To address these problems, we propose a hybrid AR architecture that combines deep learning with symbolic models to provide more realistic estimation of the classes and additional contextual information. We test the approach on a cooking dataset, describing the preparation of carrots soup. The results show that the proposed approach performs comparable to state of the art deep models inferring additional contextual properties about the current activity. The proposed approach is a first attempt to bridge the gap between deep learning and symbolic modeling for AR.","PeriodicalId":437017,"journal":{"name":"2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116972992","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-03-11DOI: 10.1109/PERCOMW.2019.8730737
I. Lendák
The primary goal of this paper is to propose and simulate crowdsourcing-based solutions which might optimize the scientific peer review system. More specifically, a global reviewer database and gamification techniques will be proposed with the goal of obtaining more high-quality reviews for papers received by journals. The proposed modifications were assessed in a multi-agent simulation environment, in which the members of the reviewer crowd were modeled as agents. Our simulation-based evaluations implemented in the MASON multi-agent environment showed that the introduction of the above improvements would allow editors to find the most suitable and responsive reviewers, as well as to lower the number of scientific papers which do not receive enough reviews.
{"title":"Simulation-based evaluation of a crowdsourced expert peer review system","authors":"I. Lendák","doi":"10.1109/PERCOMW.2019.8730737","DOIUrl":"https://doi.org/10.1109/PERCOMW.2019.8730737","url":null,"abstract":"The primary goal of this paper is to propose and simulate crowdsourcing-based solutions which might optimize the scientific peer review system. More specifically, a global reviewer database and gamification techniques will be proposed with the goal of obtaining more high-quality reviews for papers received by journals. The proposed modifications were assessed in a multi-agent simulation environment, in which the members of the reviewer crowd were modeled as agents. Our simulation-based evaluations implemented in the MASON multi-agent environment showed that the introduction of the above improvements would allow editors to find the most suitable and responsive reviewers, as well as to lower the number of scientific papers which do not receive enough reviews.","PeriodicalId":437017,"journal":{"name":"2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116870940","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-03-11DOI: 10.1109/PERCOMW.2019.8730786
Alfred Åkesson, G. Hedin, Mattias Nordahl, B. Magnusson
Future internet-of-things systems need to be able to combine heterogeneous services and support weak connectivity. In this paper, we introduce ComPOS, a new domain-specific language for composing services in IoT systems. We show how Maria, a bird watcher, can use ComPOS to build a system that allows her to spy on birds in the garden while she is not at home. We demonstrate how ComPOS handles the unpredictable nature of IoT system by analysing in what cases Maria's system is still useful when some devices are unavailable.
{"title":"ComPOS: Composing Oblivious Services","authors":"Alfred Åkesson, G. Hedin, Mattias Nordahl, B. Magnusson","doi":"10.1109/PERCOMW.2019.8730786","DOIUrl":"https://doi.org/10.1109/PERCOMW.2019.8730786","url":null,"abstract":"Future internet-of-things systems need to be able to combine heterogeneous services and support weak connectivity. In this paper, we introduce ComPOS, a new domain-specific language for composing services in IoT systems. We show how Maria, a bird watcher, can use ComPOS to build a system that allows her to spy on birds in the garden while she is not at home. We demonstrate how ComPOS handles the unpredictable nature of IoT system by analysing in what cases Maria's system is still useful when some devices are unavailable.","PeriodicalId":437017,"journal":{"name":"2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)","volume":"119 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134020461","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-03-11DOI: 10.1109/PERCOMW.2019.8730662
Wataru Sasaki, Masashi Fujiwara, Manato Fujimoto, H. Suwa, Yutaka Arakawa, K. Yasumoto
Recently, various smart home services such as smart air-conditioning, monitoring of elderly/kids and energy-efficient appliance operations are emerging, thanks to technologies of indoor positioning of users and recognition of Activity of Daily Living (ADL). Meanwhile, to realize more convenient home services, it will become more important to be able to predict occurrence time of each ADL. ADL prediction is a challenging problem since it is difficult to train a prediction model by general machine learning algorithms which use only the data at a moment for classification. In this paper, taking into account temporal dependency of data (consumed power of appliances and position of users) collected during daily life, we propose a method for constructing models to predict ADL with LSTM (Long Short-Term Memory). In the proposed method, we construct LSTM-based models by setting occurrence time of each activity to an objective variable. First, we tried to construct a multi-class classification model which outputs one of several predefined time ranges (time elapsed from present) as the occurrence time of the activity. Through preliminary experiment, we found that this model results in low accuracy in predicting the occurrence time. Then, as the second approach, we constructed a before-or-after classification model which judges if the activity occurs within a specified time or not. We applied this model to our smart home data and confirmed that it achieves better prediction accuracy for all activities.
近年来，随着用户室内定位技术和ADL (Activity of Daily Living)识别技术的发展，智能空调、老人/孩子监控、节能家电等各种智能家居服务应运而生。同时，为了实现更便捷的家庭服务，能够预测每个ADL的发生时间将变得更加重要。ADL预测是一个具有挑战性的问题，因为仅使用当前数据进行分类的一般机器学习算法难以训练预测模型。在本文中，考虑到日常生活中收集的数据(电器耗电量和用户位置)的时间依赖性，我们提出了一种基于LSTM (Long - Short-Term Memory，长短期记忆)的ADL预测模型构建方法。在该方法中，我们通过将每个活动的发生时间设置为目标变量来构建基于lstm的模型。首先，我们尝试构建一个多类分类模型，该模型输出几个预定义的时间范围(从现在开始经过的时间)中的一个作为活动的发生时间。通过初步实验，我们发现该模型对发生时间的预测精度较低。然后，作为第二种方法，我们构建了一个前后分类模型，该模型判断活动是否在指定的时间内发生。我们将这个模型应用到我们的智能家居数据中，并证实它对所有活动都有更好的预测精度。
{"title":"Predicting Occurrence Time of Daily Living Activities Through Time Series Analysis of Smart Home Data","authors":"Wataru Sasaki, Masashi Fujiwara, Manato Fujimoto, H. Suwa, Yutaka Arakawa, K. Yasumoto","doi":"10.1109/PERCOMW.2019.8730662","DOIUrl":"https://doi.org/10.1109/PERCOMW.2019.8730662","url":null,"abstract":"Recently, various smart home services such as smart air-conditioning, monitoring of elderly/kids and energy-efficient appliance operations are emerging, thanks to technologies of indoor positioning of users and recognition of Activity of Daily Living (ADL). Meanwhile, to realize more convenient home services, it will become more important to be able to predict occurrence time of each ADL. ADL prediction is a challenging problem since it is difficult to train a prediction model by general machine learning algorithms which use only the data at a moment for classification. In this paper, taking into account temporal dependency of data (consumed power of appliances and position of users) collected during daily life, we propose a method for constructing models to predict ADL with LSTM (Long Short-Term Memory). In the proposed method, we construct LSTM-based models by setting occurrence time of each activity to an objective variable. First, we tried to construct a multi-class classification model which outputs one of several predefined time ranges (time elapsed from present) as the occurrence time of the activity. Through preliminary experiment, we found that this model results in low accuracy in predicting the occurrence time. Then, as the second approach, we constructed a before-or-after classification model which judges if the activity occurs within a specified time or not. We applied this model to our smart home data and confirmed that it achieves better prediction accuracy for all activities.","PeriodicalId":437017,"journal":{"name":"2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132506424","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-03-11DOI: 10.1109/PERCOMW.2019.8730876
D. Fedotov, Yuki Matsuda, W. Minker
Recent advances in computational and sensing technologies allowed to incorporate different devices into a smart systems, making the ubiquitous or pervasive computing a hot topic for research and commercial projects. One technology, that can help the user to interact with invisible system representing smart environment is spoken dialogue system. Following the success in research on automatic speech recognition and natural language understanding, spoken dialogue systems have significantly improved themselves during the past decade and now bringing the communication between human and machine closer to natural level. Having user as a main subject, both system may benefit from explicit information about his current state and mood, adjusting their behaviour to the certain extent. In this paper we consider the combination of ubiquitous computing, spoken dialogue systems, and emotion recognition technologies, suggest possible ways of information flow, discuss future applications and potential problems. We find, that these technologies can be complementary to each other, increasing their flexibility, robustness and intelligibility when combined. We present the usage of such approach in a smart house environment, continuously tracking the state of the user, interacting with them in real time and reacting to mood changes.
{"title":"From Smart to Personal Environment: Integrating Emotion Recognition into Smart Houses","authors":"D. Fedotov, Yuki Matsuda, W. Minker","doi":"10.1109/PERCOMW.2019.8730876","DOIUrl":"https://doi.org/10.1109/PERCOMW.2019.8730876","url":null,"abstract":"Recent advances in computational and sensing technologies allowed to incorporate different devices into a smart systems, making the ubiquitous or pervasive computing a hot topic for research and commercial projects. One technology, that can help the user to interact with invisible system representing smart environment is spoken dialogue system. Following the success in research on automatic speech recognition and natural language understanding, spoken dialogue systems have significantly improved themselves during the past decade and now bringing the communication between human and machine closer to natural level. Having user as a main subject, both system may benefit from explicit information about his current state and mood, adjusting their behaviour to the certain extent. In this paper we consider the combination of ubiquitous computing, spoken dialogue systems, and emotion recognition technologies, suggest possible ways of information flow, discuss future applications and potential problems. We find, that these technologies can be complementary to each other, increasing their flexibility, robustness and intelligibility when combined. We present the usage of such approach in a smart house environment, continuously tracking the state of the user, interacting with them in real time and reacting to mood changes.","PeriodicalId":437017,"journal":{"name":"2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128534722","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-03-11DOI: 10.1109/PERCOMW.2019.8730649
Rashmi Gupta, C. Gurrin
Multimodal lifelog data consists of continual streams of multimodal sensor data about the life experience of an individual. In order to be effective, any lifelog retrieval system needs to segment continual lifelog data into manageable units. In this paper, we explore the effect of incorporating manual annotations into the lifelog event segmentation process, and we present a study into the effect of high-quality manual annotations on a query-time document segmentation process for lifelog data and evaluate the approach using an open and available test collection. We show that activity based manual annotations enhance the understanding of information retrieval and we highlight a number of potential topics of interest for the community.
{"title":"Considering Manual Annotations in Dynamic Segmentation of Multimodal Lifelog Data","authors":"Rashmi Gupta, C. Gurrin","doi":"10.1109/PERCOMW.2019.8730649","DOIUrl":"https://doi.org/10.1109/PERCOMW.2019.8730649","url":null,"abstract":"Multimodal lifelog data consists of continual streams of multimodal sensor data about the life experience of an individual. In order to be effective, any lifelog retrieval system needs to segment continual lifelog data into manageable units. In this paper, we explore the effect of incorporating manual annotations into the lifelog event segmentation process, and we present a study into the effect of high-quality manual annotations on a query-time document segmentation process for lifelog data and evaluate the approach using an open and available test collection. We show that activity based manual annotations enhance the understanding of information retrieval and we highlight a number of potential topics of interest for the community.","PeriodicalId":437017,"journal":{"name":"2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130063173","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-03-11DOI: 10.1109/PERCOMW.2019.8730730
David Schindler, Kristina Yordanova, Frank Krüger
The extraction of mentions of research artefacts from scientific papers is a necessary precursor for multiple applications ranging from simple search for literature based on particular research artefacts to semantic analyses of the investigations described in the literature. Techniques of natural language processing like named entity and relation extraction allow to establish detailed knowledge about such artefacts. The application of supervised classifiers relies on annotated datasets in order to provide a basis for training and evaluation. In this work, we present an annotation scheme for research artefacts in scientific literature which not only distinguishes between different types of artefacts like datasets, software and materials but also allows for the annotation of more detailed information such as amount or concentration of materials. Furthermore, we present first preliminary results in terms of inter-rater reliability.
{"title":"An annotation scheme for references to research artefacts in scientific publications","authors":"David Schindler, Kristina Yordanova, Frank Krüger","doi":"10.1109/PERCOMW.2019.8730730","DOIUrl":"https://doi.org/10.1109/PERCOMW.2019.8730730","url":null,"abstract":"The extraction of mentions of research artefacts from scientific papers is a necessary precursor for multiple applications ranging from simple search for literature based on particular research artefacts to semantic analyses of the investigations described in the literature. Techniques of natural language processing like named entity and relation extraction allow to establish detailed knowledge about such artefacts. The application of supervised classifiers relies on annotated datasets in order to provide a basis for training and evaluation. In this work, we present an annotation scheme for research artefacts in scientific literature which not only distinguishes between different types of artefacts like datasets, software and materials but also allows for the annotation of more detailed information such as amount or concentration of materials. Furthermore, we present first preliminary results in terms of inter-rater reliability.","PeriodicalId":437017,"journal":{"name":"2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132705623","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-03-11DOI: 10.1109/PERCOMW.2019.8730689
S. Lüdtke, Kristina Yordanova, T. Kirste
Computational causal behavior models can be used for joint human activity recognition and reasoning about the context of an activity, like the location of used objects, which is relevant for assistive systems. Such models are computationally expensive due to the large number of different states that need to be considered. However, the distribution of these states is often highly symmetrical. Lifted Marginal Filtering (LiMa) is an inference algorithm that maintains a suitably factorized state distribution, such that symmetrical factors can be represented compactly. In this paper, we show for the first time the application of LiMa to a complex real-world activity recognition setting based on real IMU data. This is achieved by introducing an operation that prevents the distribution representation to grow indefinitely, by projecting the distribution back to an exchangeable distribution. We show that LiMa needs fewer states to represent the exact filtering distribution, and achieves a higher activity recognition accuracy when only limited resources are available to represent the state distribution.
{"title":"Human Activity and Context Recognition using Lifted Marginal Filtering","authors":"S. Lüdtke, Kristina Yordanova, T. Kirste","doi":"10.1109/PERCOMW.2019.8730689","DOIUrl":"https://doi.org/10.1109/PERCOMW.2019.8730689","url":null,"abstract":"Computational causal behavior models can be used for joint human activity recognition and reasoning about the context of an activity, like the location of used objects, which is relevant for assistive systems. Such models are computationally expensive due to the large number of different states that need to be considered. However, the distribution of these states is often highly symmetrical. Lifted Marginal Filtering (LiMa) is an inference algorithm that maintains a suitably factorized state distribution, such that symmetrical factors can be represented compactly. In this paper, we show for the first time the application of LiMa to a complex real-world activity recognition setting based on real IMU data. This is achieved by introducing an operation that prevents the distribution representation to grow indefinitely, by projecting the distribution back to an exchangeable distribution. We show that LiMa needs fewer states to represent the exact filtering distribution, and achieves a higher activity recognition accuracy when only limited resources are available to represent the state distribution.","PeriodicalId":437017,"journal":{"name":"2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132920537","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：