Usman Naeem

Recognising activities of daily life using hierarchical plans

The introduction of the smart home has been seen as a way of allowing elderly people to lead an independent life for longer, making sure they remain safe and in touch with their social and care communities. The assistance could be in the form of helping with everyday tasks, e.g. notifying them when the milk in the fridge will be finished or institute safeguards to mitigate risks. In order to achieve this effectively we must know what the elderly person is doing at any given time. This paper describes a tiered approach to deal with recognition of activities that addresses the problem of missing sensor events that can occur while a task is being carried out.

Novel centroid selection approaches for KMeans-clustering based recommender systems

Recommender systems have the ability to filter unseen information for predicting whether a particular user would prefer a given item when making a choice. Over the years, this process has been dependent on robust applications of data mining and machine learning techniques, which are known to have scalability issues when being applied for recommender systems. In this paper, we propose a k-means clustering-based recommendation algorithm, which addresses the scalability issues associated with traditional recommender systems. An issue with traditional k-means clustering algorithms is that they choose the initial k centroid randomly, which leads to inaccurate recommendations and increased cost for offline training of clusters. The work in this paper highlights how centroid selection in k-means based recommender systems can improve performance as well as being cost saving. The proposed centroid selection method has the ability to exploit underlying data correlation structures, which has been proven to exhibit superior accuracy and performance in comparison to the traditional centroid selection strategies, which choose centroids randomly. The proposed approach has been validated with an extensive set of experiments based on five different datasets (from movies, books, and music domain). These experiments prove that the proposed approach provides a better quality cluster and converges quicker than existing approaches, which in turn improves accuracy of the recommendation provided.

A Comparison of Two Hidden Markov Approaches to Task Identification in the Home Environment

In todays working world the elderly are often classified as a set of dependent people and are sometimes neglected by society. One of the ways to determine whether an elderly person is safe in their home is to find out what activities an elderly person is carrying out and give appropriate assistance or institute safeguards. This paper describes the lower tier of a two tiered approach that is being adopted. The higher tier consists of hierarchical sets of plans that model common goals and sub-goals associated with activities in daily life. The lower tier deals with recognition of tasks from the stream of sensor events. Tasks are the lowest level component of a plan. The tasks are modelled using a form of hidden Markov modelling.

Activity recognition in the home using a hierarchal framework with object usage data

Smart environments are emerging as platforms that can be used to help recognise activities and hence provide context sensitive services and assistance, e.g. switching on the music while the person being monitored is having an evening meal. The ability to monitor everyday activities in a smart environment is seen as a key approach for tracking functional decline among elderly people. The motivation is to allow patients with early Alzheimers disease to have additional years of independent living before the disease reaches the latter stages (moderate and severe). This paper describes an approach to detecting the goals of the individual subjects from sensor data that are generated by objects that are used when performing everyday activities around the home. To limit intrusion into personal privacy cameras and visual surveillance equipment are not used, as the activities are monitored using simple RFID sensors. Identification of the intentions of subjects is based on interpretation of the sensor data exploiting known structures of typical behaviours.

Frequent Pattern Mining Algorithms for Finding Associated Frequent Patterns for Data Streams: A Survey☆

Abstract Pattern recognition is seen as a major challenge within the field of data mining and knowledge discovery. For the work in this paper, we have analyzed a range of widely used algorithms for finding frequent patterns with the purpose of discovering how these algorithms can be used to obtain frequent patterns over large transactional databases. This has been presented in the form of a comparative study of the following algorithms: Apriori algorithm, Frequent Pattern (FP) Growth algorithm, Rapid Association Rule Mining (RARM), ECLAT algorithm and Associated Sensor Pattern Mining of Data Stream (ASPMS) frequent pattern mining algorithms. This study also focuses on each of the algorithms strengths and weaknesses for finding patterns among large item sets in database systems.

Activity recognition using a hierarchical framework

This paper describes an approach for modelling and detecting activities of daily life based on a hierarchy of plans that contain a range of precedence relationships, representations of concurrency and other temporal relationships. Identification of activities of daily life is achieved by episode recovery models supported by using relationships expressed in the plans. The motivation is to allow people with Alzheimerpsilas disease to have additional years of independent living before the Alzheimerpsilas disease reaches the moderate and severe stages.

Recognising activities of daily life through the usage of everyday objects around the home

The integration of RFID sensors into everyday products has become a widespread solution for increasing efficiency in supply chain management. This has also led to a way of being able to monitor everyday activities in the home based on when and how these products are used, which is less intrusive than other monitoring approaches such as visual based systems. Monitoring activities in a home environment can be seen as a good way of analyzing behavior and tracking functional decline among elderly people. This paper describes a hierarchal approach for activity recognition using object usage data generated by everyday products used around the home. The motivation of this work is to allow people with early Alzheimers disease to have additional years of independent living before the disease reaches a stage where the person is fully dependable on someone else.

High-Frequency Intermuscular Coherence between Arm Muscles during Robot-Mediated Motor Adaptation

Adaptation of arm reaching in a novel force field involves co-contraction of upper limb muscles, but it is not known how the co-ordination of multiple muscle activation is orchestrated. We have used intermuscular coherence (IMC) to test whether a coherent intermuscular coupling between muscle pairs is responsible for novel patterns of activation during adaptation of reaching in a force field. Subjects (N = 16) performed reaching trials during a null force field, then during a velocity-dependent force field and then again during a null force field. Reaching trajectory error increased during early adaptation to the force-field and subsequently decreased during later adaptation. Co-contraction in the majority of all possible muscle pairs also increased during early adaptation and decreased during later adaptation. In contrast, IMC increased during later adaptation and only in a subset of muscle pairs. IMC consistently occurred in frequencies between ~40–100 Hz and during the period of arm movement, suggesting that a coherent intermuscular coupling between those muscles contributing to adaptation enable a reduction in wasteful co-contraction and energetic cost during reaching.

A Semantic Reasoning Method Towards Ontological Model for Automated Learning Analysis

Semantic reasoning can help solve the problem of regulating the evolving and static measures of knowledge at theoretical and technological levels. The technique has been proven to enhance the capability of process models by making inferences, retaining and applying what they have learned as well as discovery of new processes. The work in this paper propose a semantic rule-based approach directed towards discovering learners interaction patterns within a learning knowledge base, and then respond by making decision based on adaptive rules centred on captured user profiles. The method applies semantic rules and description logic queries to build ontology model capable of automatically computing the various learning activities within a Learning Knowledge-Base, and to check the consistency of learning object/data types. The approach is grounded on inductive and deductive logic descriptions that allows the use of a Reasoner to check that all definitions within the learning model are consistent and can also recognise which concepts that fit within each defined class. Inductive reasoning is practically applied in order to discover sets of inferred learner categories, while deductive approach is used to prove and enhance the discovered rules and logic expressions. Thus, this work applies effective reasoning methods to make inferences over a Learning Process Knowledge-Base that leads to automated discovery of learning patterns/behaviour.

Neural Correlates of Single- and Dual-Task Walking in the Real World

Recent developments in mobile brain-body imaging (MoBI) technologies have enabled studies of human locomotion where subjects are able to move freely in more ecologically valid scenarios. In this study, MoBI was employed to describe the behavioral and neurophysiological aspects of three different commonly occurring walking conditions in healthy adults. The experimental conditions were self-paced walking, walking while conversing with a friend and lastly walking while texting with a smartphone. We hypothesized that gait performance would decrease with increased cognitive demands and that condition-specific neural activation would involve condition-specific brain areas. Gait kinematics and high density electroencephalography (EEG) were recorded whilst walking around a university campus. Conditions with dual tasks were accompanied by decreased gait performance. Walking while conversing was associated with an increase of theta (θ) and beta (β) neural power in electrodes located over left-frontal and right parietal regions, whereas walking while texting was associated with a decrease of β neural power in a cluster of electrodes over the frontal-premotor and sensorimotor cortices when compared to walking whilst conversing. In conclusion, the behavioral “signatures” of common real-life activities performed outside the laboratory environment were accompanied by differing frequency-specific neural “biomarkers”. The current findings encourage the study of the neural biomarkers of disrupted gait control in neurologically impaired patients.