FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2463559436> ?p ?o ?g. }

• W2463559436 abstract "In the last decade, a new computing platform, Wireless Sensor Networks (WSN), has emerged. This platform is an interconnected and distributed transducer network of tiny, inexpensive sensors, and it has emerged as one of the essentials of contemporary ubiquitous computing. A WSN’s many tiny sensor nodes work together to perform one or more tasks, normally involving some type of monitoring, tracking and controlling. One of the main goals of WSNs is to sense physical environments and detect events occurring in the field of interest. Detection of events or materials of interest can be done by processing and analysing sensory information obtained by sensor nodes. Pattern recognition is one of the most useful and commonly utilised machine learning techniques in the literature for event detection in WSNs, especially when dealing with complex events. However, pattern recognition is highly affected by the limited resources offered by WSNs, including limited energy and limited computational, communicational and memory resources. In addition to limited resources, WSNs face other challenges in event detection, related to the dynamic nature of the environments in which WSNs are usually deployed. For example, a memorised pattern in a WSN pattern recognition scheme could appear in a different form, such as size dilation or location change, in the field of interest, or the WSN network’s topology or sensor node locations might change, meaning the information memorised within the network will have different relations and distribution. The pattern recognition scheme also requires the capability of handling noisy patterns in order to maintain a high accuracy level. These noisy patterns are mainly the result of the monitoring environment and the limited lifetime of sensor nodes. Another issue associated with the nature of WSNs is the restricted number of training instances available, as events generally occur in some form of randomness. Therefore, designing a pattern recognition scheme for event detection in WSNs is a matter of a trade-off between detection accuracy, the use of limited available resources and dealing with existing challenges. The first goal of this research project is to propose pattern recognition schemes capable of addressing the limitations associated with resource-constrained networks such as WSNs. The research first investigates the existing learning techniques for WSNs and their limitations. Then the research proposes two novel collaborative in-network pattern recognition-based event detection schemes which are lightweight and scalable and which suit resource-constrained networks such as WSNs well. In this research, two pattern recognition schemes are proposed: the Macroscopic Object Heuristics Algorithm (MOHA) and the Light Macroscopic Object Heuristics Algorithm (LMOHA). The main aim for proposing the second scheme (LMOHA) is to reduce the overall computational complexity of the MOHA scheme for event detection and pattern recognition. The proposed schemes adopt the distributed parallel recognition mechanisms of Graph Neuron (GN) to minimise recognition computations and communications and thus will lead to maintaining low levels of consumption of the limited resources. The distributed network structure of the proposed schemes will result in loosely coupled connectivity between a network’s nodes and will avoid iterative learning. Therefore, the proposed schemes will perform recognition operations in a single learning cycle of predictable duration, which will make them good candidates for implementation of large-scale, real-time problems. The second aim of this research project is to deal with a WSN’s dynamic nature and limited prior knowledge of events. Thus, pattern transformation invariant schemes are proposed in this research. The first proposed scheme (i.e. MOHA) implements an edge detection gradient-based mechanism that searches the edges and boundaries of patterns and replaces traditional local information storing. The second proposed scheme (i.e. LMOHA) implements a similar mechanism as MOHA; however, its mechanism searches for the sensory-based shapes of patterns. These mechanisms allow the proposed schemes to identify dynamic and continuous changes in patterns. Consequently, the proposed schemes will be capable of performing recognition operations in dynamic environments and will also provide a high level of detection accuracy using a minimal amount of available information about patterns. Required protocols for performing the schemes’ operations are also presented and discussed. Theoretical and experimental analysis and evaluation of the presented schemes is conducted in the research. The evaluation includes time complexity, recognition accuracy, communicational and computational overhead, energy consumption and lifetime analysis. The schemes’ performance is also compared with that of existing recognition schemes. This shows that the proposed schemes are capable of minimising computational and communicational overheads in resource-constrained networks, enabling those networks to perform efficient recognition activities for patterns that involve transformations within a single learning cycle while maintaining a high level of scalability and accuracy. The results show that a network that implements mica 2 motes and requires 3.0625 milliseconds to send a single message can perform recognition operations within a single learning cycle duration, ranging between 5.17 and 2231.39 milliseconds using the MOHA scheme and 5.17 and 16,441.33 milliseconds using the LMOHA scheme, for 40,000- and 65,536-node network settings, respectively. The results also show that using a multi-channel MAC message exchange model in both proposed schemes will considerably reduce the network’s learning cycle time. The results also show that energy requirements can be decreased by up to 75.86 per cent using the MOHA scheme and by 70.69 per cent using the LMOHA scheme, in comparison to other recognition techniques. In terms of efficiency, theoretical and experimental analyses show that both proposed schemes are highly capable of dealing with noisy and transformed patterns with a high level of accuracy. However, each presented different limits of tolerance to noisy patterns and different types of transformed patterns. The results show that the MOHA scheme offers more accurate recognition for scaled patterns than the LMOHA scheme. However, the LMOHA scheme provides more accurate recognition for noisy and rotated patterns than the MOHA scheme. In conclusion, both proposed schemes showed a very significant capability of performing pattern recognition in WSNs, as they showed a very good capability of handling noisy and transformed patterns and limiting the number of communications, and hence, limiting the use of energy resources. Finally, the research presents and discusses several simulations for each proposed scheme. The results of these simulations show that the proposed schemes have a very high accuracy in dealing with transformed patterns compared to other existing schemes. The results also show the capability of the proposed schemes’ networks of performing complex and real-life recognition problems by using a minimal amount of training information. They also show the feasibility of utilising the proposed schemes in real scenarios and different application domains." @default.
• W2463559436 created "2016-07-22" @default.
• W2463559436 creator A5049995499 @default.
• W2463559436 date "2017-03-01" @default.
• W2463559436 modified "2023-09-24" @default.
• W2463559436 title "Pattern transformation-invariant schemes for wireless sensor networks based on an edge detection gradient-based mechanism" @default.
• W2463559436 doi "https://doi.org/10.4225/03/58b6469f9c168" @default.
• W2463559436 hasPublicationYear "2017" @default.
• W2463559436 type Work @default.
• W2463559436 sameAs 2463559436 @default.
• W2463559436 citedByCount "0" @default.
• W2463559436 crossrefType "dissertation" @default.
• W2463559436 hasAuthorship W2463559436A5049995499 @default.
• W2463559436 hasConcept C104317684 @default.
• W2463559436 hasConcept C121332964 @default.
• W2463559436 hasConcept C185592680 @default.
• W2463559436 hasConcept C190470478 @default.
• W2463559436 hasConcept C204241405 @default.
• W2463559436 hasConcept C24590314 @default.
• W2463559436 hasConcept C31258907 @default.
• W2463559436 hasConcept C33923547 @default.
• W2463559436 hasConcept C37914503 @default.
• W2463559436 hasConcept C41008148 @default.
• W2463559436 hasConcept C55493867 @default.
• W2463559436 hasConcept C62520636 @default.
• W2463559436 hasConcept C89611455 @default.
• W2463559436 hasConceptScore W2463559436C104317684 @default.
• W2463559436 hasConceptScore W2463559436C121332964 @default.
• W2463559436 hasConceptScore W2463559436C185592680 @default.
• W2463559436 hasConceptScore W2463559436C190470478 @default.
• W2463559436 hasConceptScore W2463559436C204241405 @default.
• W2463559436 hasConceptScore W2463559436C24590314 @default.
• W2463559436 hasConceptScore W2463559436C31258907 @default.
• W2463559436 hasConceptScore W2463559436C33923547 @default.
• W2463559436 hasConceptScore W2463559436C37914503 @default.
• W2463559436 hasConceptScore W2463559436C41008148 @default.
• W2463559436 hasConceptScore W2463559436C55493867 @default.
• W2463559436 hasConceptScore W2463559436C62520636 @default.
• W2463559436 hasConceptScore W2463559436C89611455 @default.
• W2463559436 hasLocation W24635594361 @default.
• W2463559436 hasOpenAccess W2463559436 @default.
• W2463559436 hasPrimaryLocation W24635594361 @default.
• W2463559436 hasRelatedWork W1990012371 @default.
• W2463559436 hasRelatedWork W2116560012 @default.
• W2463559436 hasRelatedWork W2152405676 @default.
• W2463559436 hasRelatedWork W2188158166 @default.
• W2463559436 hasRelatedWork W2265318195 @default.
• W2463559436 hasRelatedWork W2280474529 @default.
• W2463559436 hasRelatedWork W2517839483 @default.
• W2463559436 hasRelatedWork W2606086581 @default.
• W2463559436 hasRelatedWork W2613057410 @default.
• W2463559436 hasRelatedWork W2869726119 @default.
• W2463559436 hasRelatedWork W2887355508 @default.
• W2463559436 hasRelatedWork W2890902148 @default.
• W2463559436 hasRelatedWork W2952379102 @default.
• W2463559436 hasRelatedWork W2960296264 @default.
• W2463559436 hasRelatedWork W3011578310 @default.
• W2463559436 hasRelatedWork W3045000028 @default.
• W2463559436 hasRelatedWork W3112480167 @default.
• W2463559436 hasRelatedWork W3166742573 @default.
• W2463559436 hasRelatedWork W3170523279 @default.
• W2463559436 hasRelatedWork W3174943183 @default.
• W2463559436 isParatext "false" @default.
• W2463559436 isRetracted "false" @default.
• W2463559436 magId "2463559436" @default.
• W2463559436 workType "dissertation" @default.