Human activity recognition (HAR), integrated with the Artificial Intelligence of Things (AIoT) technologies, enables real-time user behavior detection across a wide range of applications. However, existing multimodal HAR approaches are costly, requiring a substantial amount of labeled data and expert annotations. The baseline methods have utilized contrastive learning yet they fail to account for negative sample misclassification, and they design separate encoders per modality without considering the architectural complexity. To address these challenges, we propose contrastive fusion learning with negative set pruning (CFL-NSP), which improves upon contrastive learning-based multimodal HAR models by effectively handling misclassified negative samples. Additionally, we introduce a lightweight multi-modal feature isomorphic encoder (Li-MFIE), which unifies multimodal sensor data into a common image format, achieving improved performance over models that require separate encoders for each modality. Our experiments utilize accelerometers, gyroscope, and skeleton-based data to provide a thorough multimodal HAR analysis. Experimental results show that combining both techniques improves accuracy from 53.88% to 58.86% (+4.98%) with a 5% label rate. The CFL-NSP improves accuracy from 53.64% to 57.74% (+4.10%) by removing false negatives. On the other hand, the Li-MFIE increases accuracy from 53.88% to 55.81% (+1.93%). Additionally, using the Li-MFIE reduces FLOPs from 3775.81M to 1519.79M (-59.53%) and decreases model training time from 105s to 84s (-20.00%). These results demonstrate its potential impact on applications such as smart homes, healthcare monitoring, and fitness tracking, where real-time multimodal HAR can enhance safety and user experience. With the unified input data format, our approach has the potential to be scalable and generalizable to various HAR scenarios without additional encoder design, making it suitable for a wider range of AIoT applications.

Deploying a large number of smart cameras and training their models is a very time-consuming and labor-intensive process. Although there have been studies that utilized direct edge-to-edge (e2e) feature transfer learning without requiring help from centralized servers, they do not make good use of advanced latent features and even require constant adjustment of certain hyperparameters. This not only takes a lot of time for initial model training but also does not meet one of the key characteristics of the Internet of Thing, i.e., reduced human intervention. Therefore, we propose a Lightweight Multi-representation Attention-based Residual (LiMAR) module to reduce the computational load on an edge device via its lightweight architecture, which can extract multi-representation and diverse features to make the best use of the latent features for transfer learning in a direct e2e manner. In addition, we also propose a Lightweight Joint-distribution Autonomous Domain Adaptation (LiJADA) module to further reduce the load of model training and reduce the time required for camera deployment and the necessity of human intervention. The experimental results show that incorporating these two modules with one-to-many transfer can improve the total accuracy by 6.30% compared to a baseline study, yet with 75% less transmission data. In addition, the two modules working with many-to-one transfer learning can reduce the transmission cost by 83.33% and improve the total accuracy by 5.10%.Note to Practitioners—This study was motivated by the practical problem of enabling an unmanned application to efficiently learn its initial models. This problem is important for practitioners in unmanned areas who aim to reduce the effort required for system deployment by utilizing direct edge-device to edge-device (or edge-to-edge) model transfer learning. Existing approaches mainly rely on basic feature-level transfer learning, and they often require constant adjustment of hyperparameters to achieve optimal model performance. To tackle these challenges, we propose a novel approach that involves transferring multi-representation features with automatic hyper-parameter adjustment. This approach can improve knowledge reusability across different environments, and it is also important for real-life scenarios. Preliminary evaluations suggest that the approach is feasible and can achieve knowledge reuse while reducing data transmission costs. However, since we still use raw input images for matching the source and target cameras rather than using features for matching, this may limit non-source-free applications. To address this limitation, incorporating model-based transfer in the next-phase implementation can mitigate the drawback.

In recent years, Artificial Intelligence of Things (AIoT) has been widely adopted across various smart systems, accelerating the development of edge computing. Nevertheless, existing research on end-to-end dense video captioning falls short in one of two areas: 1) either it does not prioritize global information or 2) it tends to focus on irrelevant details. Our study proposes an end-to-end dense video captioning model with sequence-aware learnable sparse mask. This model results in improved focus on essential information in a video while ignoring irrelevant details, thus enhancing the quality of caption generation. In addition, existing video captioning research which uses all input video frames are frequently hampered by redundancy and thus generate incorrect captions. To overcome this issue, we propose a lightweight frame selection model that primarily utilizes our proposed lightweight attention-enhancement residual gated network to achieve the desired accuracy with a smaller computational cost. The effectiveness of our proposed approaches was tested and compared to existing models. Our model achieved higher accuracy compared to previous studies, and the lightweight frame selection network resulted in higher efficiency while generating more accurate captions after frame selection.

With the gradual advancement of technology in the production of consumer goods, the Internet-of-Things (IoT) systems have experienced rapid development, resulting in a massive amount of data that can be processed using deep neural networks. However, annotating the data and training the models require significant manpower, time, and computational resources. Transfer learning can address this problem. Traditional systems rely on centralized servers for transfer learning. Although several studies have proposed distributed systems for direct edge-to-edge instance-based and feature-based transfer learning, they neglect model-based transfer learning within the same domain. This leads to lower learning efficiency, lower privacy protection, and higher transmission costs. Therefore, our study proposes direct edge-to-edge local-learning-assisted model-based transfer learning for a direct edge-to-edge many-to-many model-based transfer learning scenario. The method can transfer model structures and weights between distributed devices without relying on powerful centralized servers. The effectiveness of the proposed approach is demonstrated by applying it to various scenarios.

Recently, AI has gradually been integrated into various transportation tools. Additionally, precise environmental perception is key to ensuring the stable operation of autonomous vehicles. Although current research has started to focus on the processing of various environmental perception tasks with point-cloud data, existing studies can only achieve a “box-level” hybrid scene understanding, with limitations in recognizing more fine-grained “point-level” variations. Therefore, we propose a multitask network with hierarchical channel-attention gating” that can flexibly and simultaneously handle 3-D semantic segmentation and scene flow estimation within a single architecture to provide point-level hybrid scene understanding. In addition, we also identify the issue of underutilization of training data in existing studies. Due to the difficulty of point-cloud data labeling, some point-cloud data are left unlabeled to reduce the cost of labeling, but this also results in the underutilization of the unlabeled data. To efficiently incorporate unlabeled data into the learning process, we propose a strategy involving point-level self-training for a multitask network. This technique employs flow information to generate reliable pseudo-labels for unlabeled data, before integrating them into the training process to enhance model performance. The experimental outcomes on the Waymo open data set demonstrate the superiority of the proposed network.

In recent years, the Artificial Intelligence of Things (AIoT) has led to the rapid development of edge computing, and existing video-captioning systems can be deployed directly on AIoT-enabled cameras (hereafter referred to as edge cameras), which have increasingly powerful computing resources. Therefore, we propose a lightweight dense-video-captioning model based on the Transformer framework to improve execution efficiency for video-caption generation on edge cameras. In addition, to investigate the effect of concept drift on video captioning, we also propose an environment-aware adaptation to allow the system to respond to changes in the environment in order to produce more accurate video captions. Experimental results on the lightweight dense-video-captioning model show that the bilingual evaluation understudy (BLEU) metrics can increase by up to 23.5%, the computation time decreases by 46.4%, and the edge camera operates at a rate of 27.63 FPS, which is also faster than existing approaches by 4.7%. In addition, the mean average precision of the environment-aware adaptation is up to 11.3% higher than the existing approaches. In conclusion, the proposed approaches perform better than previous ones and are more flexible in different weather scenarios.

Smart cameras that leverage edge-computing (edge cameras) are increasingly being combined with deep neural networks to realize the Artificial Intelligence of Things (AIoT), enabling a smart life with “low-touch services,” such as unmanned stores. However, deploying many edge cameras and training their models (edge models) in unmanned stores is time consuming and labor intensive. Therefore, studies have utilized transfer learning methods, but training edge models often need powerful servers. Although we previously proposed direct edge-to-edge instance transfer, this approach did not exploit latent features, still required high bandwidth and long training time, and caused privacy leakage. Therefore, we propose direct edge-to-edge many-to-many latent feature transfer learning, which includes elite latent feature (ELF) extraction, direct edge-to-edge one-to-many latent feature transfer learning (DeOmf), and direct edge-to-edge many-to-one latent feature transfer learning (DeMof). Through ELF extraction, DeOmf allows one source edge camera to transfer latent features to multiple target edge cameras, which improves knowledge reuse and accelerates initial model training. Further, DeMof can help exploit the diversity of multiple source edge cameras for one target edge camera. The experimental results show that DeOmf improves accuracy by 6.30%, reduces training time by 32.15%, and saves 22.92% (maximum 83.33%) on transmission cost. In addition, DeMof increases accuracy by 3.42% and saves 66.99% on training time and 56.67% on transmission cost. These improvements can greatly broaden the applicability of our proposed system when comparing to instance-based transfer learning.

Blurred images often significantly influence the stability of computer vision systems. With the development of the Internet of Things (IoT), a camera that leverages artificial intelligence and edge computing (hereafter referred to as an edge camera) can enhance the robustness of an IoT service. Although previous studies have used deep neural networks (DNNs) for image deblurring, existing image-deblurring techniques result in over-smoothing, even with a lightweight model design. In addition, previous deblurring studies have not considered the quality of the input image first; thus, a clear input image would be processed with unnecessary deblurring. Therefore, our study proposes a lightweight and dynamic border-enhancement deblurring deep network that can be operated with high performance on edge cameras, and the network can assess the quality of input images to determine whether deblurring is necessary. The experimental results show that our lightweight deblurring approach outperforms existing studies by up to 8%, given that the speed also improved by 33.78%. As for the dynamic deblurring approach, our results show that the image metrics can be slightly improved and that speed is improved by 294.5%.

Although deep neural network (DNN) models have been extensively studied, they are often too complex to directly execute in real time on smart cameras with limited resources. Time-domain (TD) DNN models that incorporate frequency-domain (FD) layers, also known as hybrid-domain models, have been developed to mitigate the above problem; however, these require time/FD transforms. Currently, a pure FD DNN does not exist. Thus, our study proposes the first of such, along with our lightweight time/FD transform. Our model ensures that the networks perform faster on smart cameras and are more memory and energy efficient, both of which are important for smart cameras utilizing edge computing. Unlike existing TD or hybrid-domain studies, our model optimizes several internal neural network layers and implements a lightweight time/FD transform to reduce the number of calculations. More importantly, our study is the first to realize an FD fully connected layer, which can better represent a spectral feature distribution. The experimental results show that our accuracy slightly outperforms that of the existing time and hybrid-domain studies. In addition, our model’s inference speed on the edge computing platform was shown to be faster by a maximum of 52.01% for the MNIST data set and 52.00% for the CIFAR-10 data set. Furthermore, our model can improve frames per second (FPS) by at least 52.00% and memory usage by 43.64%, and save approximately 26.09% of power consumption for the MNIST data set.

Due to advances in edge computing, image recognition via smart cameras (hereafter referred to as edge cameras) has facilitated the development of unmanned stores. However, it is very time consuming and costly to collect labeled data for training initial models for an edge camera. Although existing transfer learning can speed up model training, it must depend on a powerful centralized server with considerable human intervention, thus hindering the development of autonomous and collaborative edge learning. To address this issue, our study proposes direct edge-to-edge (e2e) collaborative transfer learning with three key technologies. The first is elite-instance-based matching to utilize and transmit only representative images for transfer learning to decrease network cost among edge cameras. The second is one-to-many e2e transfer learning, which can increase the knowledge reusability of a single source camera to build multiple target models. The last is many-to-one e2e transfer learning, which enables a target edge camera to reuse knowledge from multiple sources to further decrease the effort in labeled data collection. The experimental results show that elite-instance-based matching can effectively save up to 70% of source samples on average that need to be transmitted for initial model training, and it further improves the accuracy of existing one-to-many and many-to-one e2e transfer learning.

For the upcoming Internet of Things (IoT) enabled era, context-aware service provisioning (CaSP) can be realized by first analyzing new input data, followed by inferring contexts from the input data, and providing new services based on the inferred contexts. Over time, further new contextual models will also be incorporated into CaSP due to growing data collected by existing or new IoT devices. Furthermore, these contextual models require ongoing adaptation because a real-life environment is dynamic in nature. It becomes more challenging to maintain and adapt these ever-increasing contextual models as the system evolves. To address these concerns, this paper proposes a CaSP infrastructure along with an agentized and reconfigurable design to improve system adaptability and extensibility. The proposed middleware-enhanced CaSP infrastructure can keep as much previously learned knowledge as possible to share among all integrated components. This design reduces the overhead from integrating and adapting multiple contextual models in response to inevitable uncertainties in a dynamically changing IoT-enabled smart home environment. Our agentized design generalizes the scheme of all smart components integrated with the CaSP infrastructure, thus facilitating reciprocal cooperation among all initially independent components. The CaSP infrastructure can facilitate multilevel rather than single-level information reuse via message queues residing in the middleware. This reconfigurable design enables all of the smart components to become loosely coupled and flexibly interconnect to form reconfigurable agents. Such a design allows temporal reconfiguration by reusing as many existing or even new features and contexts on the CaSP infrastructure, thus improving extensibility. Our experimental results show that the proposed CaSP infrastructure can improve the overall adaptability (by about 20%) and extensibility (by 44% to 95%) of CaSP in a dynamic environment.

The customization of avatars can help students immerse themselves in game-based learning. However, different individuals have distinct characteristics, especially game experience (GE) and cognitive styles, which may lead to different preferences for the customization of avatars. Thus, this study aims to investigate how GE and cognitive styles affect students’ reactions toward customizable avatars. Two studies, quantitative and qualitative, were conducted for system evaluation. A total of 82 students participated in Study One, where they interacted with both a customizable avatar and an ordinary avatar. The findings from Study One indicated that the students using the customizable version experienced a stronger sense of presence and flow experience than those who used the ordinary version. Regarding GE, the low GE students showed an enhanced sense of presence whereas the high GE students expressed deeper engagement. Regarding cognitive styles, Pask’s Holism/Serialism was adopted. Holists experienced an enhanced feeling of presence whereas Serialists showed deeper engagement. On the other hand, Study Two was conducted with a qualitative approach, where 11 students were further interviewed. The results showed that GE considerably affected their reactions, in terms of favored preferences and engagement, whereas cognitive styles did not have great effects. Based on the findings, a design framework was proposed for the development of personalized game-based learning systems in the future.

Home energy savings via eco-feedback is often considered a serious, tedious, or even distracting task due to the need for frequent human intervention. In addition, most existing eco-feedback systems focus on providing energy usage information and ignore the users' contexts. This may render the systems incapable of capturing the users' changes when proenvironmental behaviors have been encouraged. To address this, our study leverages Internet of Things (IoT) enabled technologies to realize an adaptive, context-aware, and playful cyber-physical system (CPS) for everyday energy savings with the hope of facilitating collaboration between a person and a smart energy-saving (ES) system to the greatest extent possible. To leverage the benefits inherent in a CPS, bidirectionally interactive information visualization integrated with pet-raising gamification was incorporated into the adaptive CPS. This synchronized the information from the user's physical environment with its counterpart in the pet's virtual environment. In order to provide flexible services, all IoT devices were agentized to form reconfigurable agents. In our experimental evaluation, this bidirectional mapping empowered users to flexibly control remote appliances anywhere and anytime in a more natural way, which enabled users to embed interactions with the ES CPS into their daily routine. Furthermore, improvements to system adaptability (33% in precision; 21% in recall) along with the reconfigurable ES services (additionally saving energy about 16.21%) show the potential of the CPS to enhance the users' experience and prolong the users' engagement in everyday energy savings.

Most previous work in household energy conservation has focused on rule-based home automation to achieve energy savings, with relatively few researchers focusing on context-aware technologies. As a result, user comfort is often disregarded and few solutions handle decision conflicts caused by multiple activities undertaken by multiple users. The main contribution of this work is twofold. First, a comprehensive human-centric and context-aware comfort index is proposed to evaluate how users feel under particular environmental conditions with regard to thermal, illumination, and appliance-usage preferences. Second, the energy savings is formulated into an optimization problem to minimize the total energy consumption, even under multiple user comfort constraints. Short-term evaluation in our simulated home environment resulted in energy savings of at least 28.98%. Long-term evaluation using a home simulator resulted in energy savings of 33.7%. Most importantly, the energy savings in both situations was achieved under multiple user comfort constraints, representing a truly human-centric living environment.

Building contextual models for new “smart” environments is not considered cost effective if data for model training must be collected from scratch. It is more practical to transfer as much learned knowledge as possible from an existing environment to the new target environment in order to reduce the data collection effort. In order to reuse learned knowledge from an original environment, this study proposed a feature-based knowledge transfer framework. The framework makes use of transfer learning, which relaxes the constraint requiring model training and testing datasets to be highly similar in distribution. Experimental results show that this framework can successfully help extract and transfer knowledge between two different smart-home environments. Models trained via the proposed framework can even outperform nontransfer-learning models by up to 8% in accuracy. Finally, the flexibility of the proposed framework enables used as a test bed for evaluating different methods and models in order to improve the service quality of human-centric context-aware applications.

Activity recognition (AR) is a key enabler for a context-aware smart home since knowing what the residents’ current activities helps a smart home provide more desirable services. This is why AR is often used in assistive technologies for cognitively impaired people to evaluate their abilities to undertake activities of daily living. In a real-life scenario, multiple-resident AR has been considered as a very challenging problem, primarily due to the complexity of data association. In addition, most prior research has not considered the potential interpersonal interactions among residents to simplify complexity, especially in an environment monitored by ambient sensors. In this study, we propose two types of multiuser activity models, both of which are derived from an interaction-feature enhanced multiuser model learning framework. These two models consider interpersonal interactions and data association for multiuser AR using ambient sensors. We then compare their performance with the other two baseline models with or without consideration of data association and interpersonal interactions. The experimental results show that the derived models outperform other baseline classifiers. Therefore, the proposed approach can increase the opportunities for providing context-aware services for a multiresident smart home.

Human activity is among the critical information for a context-aware energy saving system since knowing what activities are undertaken is important for judging if energy is well spent. Most of the prior works on energy saving do not make the best of context-awareness especially in a multiuser environment to assist the energy saving system. In addition, they often ignore whether appliances are operating implicitly or explicitly related to the context. These factors may compromise the practicality and acceptability of most of the currently available energy saving systems, thus failing to meet real user needs. Therefore, we propose Energy-Responsive Aggregate Context (ERAC) to model multi-resident activities and their associated energy consumption. Based on the relationship, implicit or explicit, between a given appliance and its associated context, an energy saving system and its users can better determine whether the power consumed by the appliance is wasted. Our experimental results demonstrate the effectiveness of the proposed approach.

Identifying on-going activities for the provision of services that are capable of matching the needs of users poses a number of daunting challenges. Most existing approaches to activity recognition require training offline activity models before being applied to the identification of activities in real time. However, the dynamic nature of actual living environments can make previously learned activity models irrelevant. This study addressed the problem of learning and recognizing daily activities in a dynamic smart-home environment, using a novel approach referred to as hybrid user-assisted incremental model adaptation. This approach involves reconfiguring previously learned activity models within a dynamic environment, while pursuing maximum efficiency by using assistance from users as well as the system to annotate new training data. Experiments that are conducted in a fully equipped smart-home lab demonstrate the efficacy of the proposed approach.

Energy overuse has caused many environmental and economic issues, so energy saving for household is challenging and important for a smart home. For home energy saving based on context-awareness, human activity is critical information since knowing what activities are undertaken is important for judging if energy consumed by appliances is well spent by users. Such contextual information is an important clue for providing an energy saving service. However, most of the prior works on home energy saving often ignore those appliances which are operating indirectly or implicitly related to the context. These factors may compromise the practicality and acceptability of most of the currently available energy saving systems, thus failing to meet real user needs. Therefore, we propose utilizing an Energy-Prone Context to model a context and its associated energy consumption. In addition, we also propose a systematic method to determine energy-saving services based on the Energy-Prone Contexts. Our experimental results demonstrate the effectiveness of the proposed approach.

Human activity is among the critical information for a context-aware smart home since knowing what activities are undertaken is important for providing appropriate services. Most of the prior works primarily focus on recognizing individual activity, thus requiring high cost to track people and performs not well when there are multiple users, which is common in a real home environment. Therefore, we propose hierarchical generalized context inference to infer multi-user contexts. By treating a multi-user context as a generalized context caused by an aggregated entity, our approach generalizes these multi-user contexts with different information granularity, and then dynamically infers and aggregates these generalized contexts. Based on the inference results of generalized contexts, a context-aware smart home can provide appropriate services as much as possible. Our experimental results demonstrate the effectiveness of the proposed approach.

Recognizing Activities of Daily Living (ADL) plays an important role in healthcare. However, it is often impractical and sometimes impossible for a person to collect those useful data manually, not to mention constant long-term data maintenance and analysis. To address the above-mentioned challenges, we propose an architecture, in which many health-care applications and services can easily build upon, for collective long-term ADL pattern analysis that leverages several prominent advantages inherent in cloud computing. The core of the proposed infrastructure includes a module to perform MapReduce-assisted Bayesian activity recognition based on all collected ADL data. Better yet, the resultant data analysis can be delivered as a service from a service station which serves as a readily accessible interface to 3rdparty service providers and end-users. For the evaluation of the proposed architecture, a simulation of persuasive health engagement is presented and discussed as one potential application.

Energy saving has become an important issue in recent years due to the problems relating to energy shortage and global warming. However, most prior works in a domestic environment often ignore users' perception of the deployed technologies, let alone users' high-level contexts such as on-going activities and preferences. Without the high-level contexts, an ES system may fail to provide sufficient clues for a user to determine the most desirable ES strategy. In addition, the prior works are more technology-oriented and assume that our real-life environment stayed fixed once the system has been established. This causes their energy-saving (ES) strategies to be less human-centric and less adaptive to users' context changes. Moreover, good ES strategies are often established after the collection of long-term data and thorough analysis such that efficient ES models can be well trained, but such models are often not reusable or not sharable among different users or even communities. This indirectly leads to extra "en-ergy" waste in setting up a new energy-efficient environment. To remedy these drawbacks, in this paper we propose a cloud-enabled adaptive activity-aware energy-saving system which not only can facilitate more human-centric and context-aware ES strategies in a dynamic environment, but also can share the promising ES models that embed these desirable ES strategies among different users to facilitate community-based energy-saving.

Every home has its own unique considerations for location-aware applications. This makes a flexible architecture very crucial for efficiently integrating various tracking devices/models for adapting to real human needs. Here, we propose a reciprocal and extensible architecture to flexibly add/remove tracking sensors/models for tracking multiple targets in a smart home. Regarding tracking devices, we employ sensors from two different categories, those with seamless sensors and those with seamful ones. This allows us to take human-centric needs into consideration and to facilitate reciprocal and cooperative interaction among sensors from the two categories. Such reciprocal cooperation aims to increase the accuracy of location estimates and to compensate for the limitations of each sensor or a tracking algorithm, which allows us to track multiple targets simultaneously in a more reliable way. Moreover, the approach demonstrated in this paper can serve as a guideline to help users customize sensor arrangements to fulfill their requirements. Our experimental results, which comprise three tracking scenarios using a load sensory floor as the seamless sensor and RF identifications (RFIDs) as seamful sensors, demonstrate the effectiveness of the proposed architecture.

Multi-resident activity recognition is among a key enabler in many context-aware applications in a smart home. However, most of prior researches ignore the potential interactions among residents in order to simplify problem complexity. On the other hand, multiple-resident activities are usually recognized using cameras or wearable sensors. However, due to human-centric concerns, it is more preferable to avoid using obtrusive sensors. In this paper, we propose dynamic Bayesian networks which extend coupled hidden Markov models (CHMMs) by adding some vertices to model both individual and cooperative activities. In order to improve performance of the model, we categorize sensor observations based on data association and some domain knowledge to model multiple-resident activity patterns. We then validate the performance using a multi-resident dataset from WSU (Washington State University), which only includes non-obtrusive sensors. The experimental result shows that our model performs better than other baseline classifiers.

It is desirable to know a resident's on-going activities before a robot or a smart system can provide attentive services to meet real human needs. This work addresses the problem of learning and recognizing human daily activities in a dynamic environment. Most currently available approaches learn offline activity models and recognize activities of interest on a real time basis. However, the activity models become outdated when human behaviors or device deployment have changed. It is a tedious and error-prone job to recollect data for retraining the activity models. In such a case, it is important to adapt the learnt activity models to the changes without much human supervision. In this work, we present a self-reconfigurable approach for activity recognition which reconfigures previously learnt activity models and infers multiple activities under a dynamic environment meanwhile pursuing minimal human efforts in relabeling training data by utilizing active-learning assistance.

Reliable recognition of activities from cluttered sensory data is challenging and important for a smart home to enable various activity-aware applications. In addition, understanding a user's preferences and then providing corresponding services is substantial in a smart home environment. Traditionally, activity recognition and preference learning were dealt with separately. In this work, we aim to develop a hybrid system which is the first trial to model the relationship between an activity model and a preference model so that the resultant hybrid model enables a preference model to assist in recovering performance of activity recognition in a dynamic environment. More specifically, on-going activity which a user performs in this work is regarded as high level contexts to assist in building a user's preference model. Based on the learned preference model, the smart home system provides more appropriate services to a user so that the hybrid system can better interact with the user and, more importantly, gain his/her feedback. The feedback is used to detect if there is any change in human behavior or sensor deployment such that the system can adjust the preference model and the activity model in response to the change. Finally, the experimental results confirm the effectiveness of the proposed approach.

This paper presents a robust location-aware activity recognition approach for establishing ambient intelligence applications in a smart home. With observations from a variety of multimodal and unobtrusive wireless sensors seamlessly integrated into ambient-intelligence compliant objects (AICOs), the approach infers a single resident's interleaved activities by utilizing a generalized and enhanced Bayesian Network fusion engine with inputs from a set of the most informative features. These features are collected by ranking their usefulness in estimating activities of interest. Additionally, each feature reckons its corresponding reliability to control its contribution in cases of possible device failure, therefore making the system more tolerant to inevitable device failure or interference commonly encountered in a wireless sensor network, and thus improving overall robustness. This work is part of an interdisciplinary Attentive Home pilot project with the goal of fulfilling real human needs by utilizing context-aware attentive services. We have also created a novel application called ldquoActivity Maprdquo to graphically display ambient-intelligence-related contextual information gathered from both humans and the environment in a more convenient and user-accessible way. All experiments were conducted in an instrumented living lab and their results demonstrate the effectiveness of the system.

Weaving computing technologies into a living environment without interfering with natural interactions is nontrivial. In this paper, we have proposed utilizing ambient-intelligence compliant object (AICO) to facilitate context-aware service provision in a smart home; furthermore, a hybrid weaving strategy, called Hide and Not Easy to Seek, is proposed for designing the weaving layer of each AICO and for popularizing ubiquitous computing. Seven weaving guidelines which combine seamless and seamful designs are learned from our actual instrumentation of a living lab and continuous cooperation with specialists from various domains. By following the proposed guidelines, our expectation is that people will be able to use the weaved technologies more naturally to accomplish everyday tasks.

Here we propose an extensible multiple target tracking architecture which utilizes RFID (Radio Frequency Identification) technology and cooperates with currently existing tracking sensors. Our proposed approach demands every individual sensor to predefine a confidence factor in advance, which allows us to track multiple targets simultaneously in a more accurate and power-efficient way. We have previously shown that a load sensory floor can non-intrusively keep track of residents' locations based on both their regular movement patterns and distinguishable features such as their weights; however, it generally becomes infeasible to simultaneously differentiate two or more residents with irregular movement patterns or similar weight measurements. With the assistance of active RFID tags, our proposed system can cope with these limitations and improve upon the tracking results of previous systems by fusing RFID-based correction signals with those from currently available sensors. Each sensor has to reckon its corresponding confidence factor which is later used in conjunction with the RFID signals to dynamically discriminate erroneous outputs from correct ones, meanwhile mitigating the power consumption issues inherent in RFID systems. Our experimental results, which comprise three scenarios having distinct irregular movement patterns, demonstrate the effectiveness of the architecture.

The rapid advancement in computer technology enables home automation system to provide a variety of convenient and novel services to people. Generally speaking, locating residents' positions in home environment is a key issue for service provision. In this paper, we propose a human localization system for our Smart Home. The human localization system uses the Condensation algorithm to locate residents' positions via multi-camera and sensory floor approaches. The Condensation algorithm is a kind of Bayesian filters and has ability to handle multi-target tracking. By integrating information from multiple sensors, we can overcome the static occlusion in the environment and make the localization system more robust. We also describe architecture of the proposed and discuss its performance.