Elliptical summary randomisation for sensor-based human activity recognition

Many conventional statistical machine learning al- gorithms generalise poorly if distribution bias ex- ists in the datasets. For example, distribution bias arises in the context of domain generalisation, where knowledge acquired from multiple source domains need to be used in a previously unseen target domains. We propose Elliptical Summary Randomisation (ESRand), an efficient domain generalisation approach that comprises of a randomised kernel and elliptical data summarisation. ESRand learns a domain interdependent projection to a la- tent subspace that minimises the existing biases to the data while maintaining the functional relationship between domains. In the latent subspace, ellipsoidal summaries replace the samples to enhance the generalisation by further removing bias and noise in the data. Moreover, the summarisation enables large-scale data processing by significantly reducing the size of the data. Through comprehensive analysis, we show that our subspace-based approach outperforms state-of-the-art results on several activity recognition benchmark datasets, while keeping the computational complexity significantly low.

Machine learning for activity recognition : hip versus wrist data

Problem addressed Wrist-worn accelerometers are associated with greater compliance. However, validated algorithms for predicting activity type from wrist-worn accelerometer data are lacking. This study compared the activity recognition rates of an activity classifier trained on acceleration signal collected on the wrist and hip. Methodology 52 children and adolescents (mean age 13.7 +/- 3.1 year) completed 12 activity trials that were categorized into 7 activity classes: lying down, sitting, standing, walking, running, basketball, and dancing. During each trial, participants wore an ActiGraph GT3X+ tri-axial accelerometer on the right hip and the non-dominant wrist. Features were extracted from 10-s windows and inputted into a regularized logistic regression model using R (Glmnet + L1). Results Classification accuracy for the hip and wrist was 91.0% +/- 3.1% and 88.4% +/- 3.0%, respectively. The hip model exhibited excellent classification accuracy for sitting (91.3%), standing (95.8%), walking (95.8%), and running (96.8%); acceptable classification accuracy for lying down (88.3%) and basketball (81.9%); and modest accuracy for dance (64.1%). The wrist model exhibited excellent classification accuracy for sitting (93.0%), standing (91.7%), and walking (95.8%); acceptable classification accuracy for basketball (86.0%); and modest accuracy for running (78.8%), lying down (74.6%) and dance (69.4%). Potential Impact Both the hip and wrist algorithms achieved acceptable classification accuracy, allowing researchers to use either placement for activity recognition.

Enhancing human action recognition with region proposals

Deep convolutional network models have dominated recent work in human action recognition as well as image classification. However, these methods are often unduly influenced by the image background, learning and exploiting the presence of cues in typical computer vision datasets. For unbiased robotics applications, the degree of variation and novelty in action backgrounds is far greater than in computer vision datasets. To address this challenge, we propose an “action region proposal” method that, informed by optical flow, extracts image regions likely to contain actions for input into the network both during training and testing. In a range of experiments, we demonstrate that manually segmenting the background is not enough; but through active action region proposals during training and testing, state-of-the-art or better performance can be achieved on individual spatial and temporal video components. Finally, we show by focusing attention through action region proposals, we can further improve upon the existing state-of-the-art in spatio-temporally fused action recognition performance.

Human activity-induced vibration in slender structural systems

Human activity-induced vibrations in slender structural sys tems become apparent in many different excitation modes and consequent action effects that cause discomfort to occupants, crowd panic and damage to public infrastructure. Resulting loss of public confidence in safety of structures, economic losses, cost of retrofit and repairs can be significant. Advanced computational and visualisation techniques enable engineers and architects to evolve bold and innovative structural forms, very often without precedence. New composite and hybrid materials that are making their presence in structural systems lack historical evidence of satisfactory performance over anticipated design life. These structural systems are susceptible to multi-modal and coupled excitation that are very complex and have inadequate design guidance in the present codes and good practice guides. Many incidents of amplified resonant response have been reported in buildings, footbridges, stadia a nd other crowded structures with adverse consequences. As a result, attenuation of human-induced vibration of innovative and slender structural systems very ofte n requires special studies during the design process. Dynamic activities possess variable characteristics and thereby induce complex responses in structures that are sensitive to parametric variations. Rigorous analytical techniques are available for investigation of such complex actions and responses to produce acceptable performance in structural systems. This paper presents an overview and a critique of existing code provisions for human-induced vibration followed by studies on the performance of three contrasting structural systems that exhibit complex vibration. The dynamic responses of these systems under human-induced vibrations have been carried out using experimentally validated computer simulation techniques. The outcomes of these studies will have engineering applications for safe and sustainable structures and a basis for developing design guidance.

Physical activity recognition from accelerometer data using a multi-scale ensemble method

Accurate and detailed measurement of an individual's physical activity is a key requirement for helping researchers understand the relationship between physical activity and health. Accelerometers have become the method of choice for measuring physical activity due to their small size, low cost, convenience and their ability to provide objective information about physical activity. However, interpreting accelerometer data once it has been collected can be challenging. In this work, we applied machine learning algorithms to the task of physical activity recognition from triaxial accelerometer data. We employed a simple but effective approach of dividing the accelerometer data into short non-overlapping windows, converting each window into a feature vector, and treating each feature vector as an i.i.d training instance for a supervised learning algorithm. In addition, we improved on this simple approach with a multi-scale ensemble method that did not need to commit to a single window size and was able to leverage the fact that physical activities produced time series with repetitive patterns and discriminative features for physical activity occurred at different temporal scales.

Multiview gait biometrics for human identity recognition

We propose a novel multiview fusion scheme for recognizing human identity based on gait biometric data. The gait biometric data is acquired from video surveillance datasets from multiple cameras. Experiments on publicly available CASIA dataset show the potential of proposed scheme based on fusion towards development and implementation of automatic identity recognition systems.

Activity recognition using binary tree SVM

This paper presents an effective classification method based on Support Vector Machines (SVM) in the context of activity recognition. Local features that capture both spatial and temporal information in activity videos have made significant progress recently. Efficient and effective features, feature representation and classification plays a crucial role in activity recognition. For classification, SVMs are popularly used because of their simplicity and efficiency; however the common multi-class SVM approaches applied suffer from limitations including having easily confused classes and been computationally inefficient. We propose using a binary tree SVM to address the shortcomings of multi-class SVMs in activity recognition. We proposed constructing a binary tree using Gaussian Mixture Models (GMM), where activities are repeatedly allocated to subnodes until every new created node contains only one activity. Then, for each internal node a separate SVM is learned to classify activities, which significantly reduces the training time and increases the speed of testing compared to popular the `one-against-the-rest' multi-class SVM classifier. Experiments carried out on the challenging and complex Hollywood dataset demonstrates comparable performance over the baseline bag-of-features method.

Human action recognition from video sequences

This PhD research has proposed new machine learning techniques to improve human action recognition based on local features. Several novel video representation and classification techniques have been proposed to increase the performance with lower computational complexity. The major contributions are the construction of new feature representation techniques, based on advanced machine learning techniques such as multiple instance dictionary learning, Latent Dirichlet Allocation (LDA) and Sparse coding. A Binary-tree based classification technique was also proposed to deal with large amounts of action categories. These techniques are not only improving the classification accuracy with constrained computational resources but are also robust to challenging environmental conditions. These developed techniques can be easily extended to a wide range of video applications to provide near real-time performance.

Supervised latent dirichlet allocation models for efficient activity representation

Local spatio-temporal features with a Bag-of-visual words model is a popular approach used in human action recognition. Bag-of-features methods suffer from several challenges such as extracting appropriate appearance and motion features from videos, converting extracted features appropriate for classification and designing a suitable classification framework. In this paper we address the problem of efficiently representing the extracted features for classification to improve the overall performance. We introduce two generative supervised topic models, maximum entropy discrimination LDA (MedLDA) and class- specific simplex LDA (css-LDA), to encode the raw features suitable for discriminative SVM based classification. Unsupervised LDA models disconnect topic discovery from the classification task, hence yield poor results compared to the baseline Bag-of-words framework. On the other hand supervised LDA techniques learn the topic structure by considering the class labels and improve the recognition accuracy significantly. MedLDA maximizes likelihood and within class margins using max-margin techniques and yields a sparse highly discriminative topic structure; while in css-LDA separate class specific topics are learned instead of common set of topics across the entire dataset. In our representation first topics are learned and then each video is represented as a topic proportion vector, i.e. it can be comparable to a histogram of topics. Finally SVM classification is done on the learned topic proportion vector. We demonstrate the efficiency of the above two representation techniques through the experiments carried out in two popular datasets. Experimental results demonstrate significantly improved performance compared to the baseline Bag-of-features framework which uses kmeans to construct histogram of words from the feature vectors.

An activity theory framework for industrial design

Ethnography has gained wide acceptance in the industrial design profession and curriculum as a means of understanding the user. However, there is considerable confusion about the particularities of its practice accompanied by the absence of an interoperable vocabulary. The consequent interdisciplinary effort is a power play between disciplines whereby the methodological view of ethnography marginalises its theoretical and analytical components. In doing so, it restricts the potential of ethnography suggesting the need for alternative methods of informing the design process. This article suggests that activity theory, with an emphasis on human activity as the fundamental unit of study, is an appropriate methodology for the generation of user requirements. The process is illustrated through the adaptation of an ethnographic case study, for the design of classroom furniture in India.

Simple classification of walking activities using commodity smart phones

People interact with mobile computing devices everywhere, while sitting, walking, running or even driving. Adapting the interface to suit these contexts is important, thus this paper proposes a simple human activity classification system. Our approach uses a vector magnitude recognition technique to detect and classify when a person is stationary (or not walking), casually walking, or jogging, without any prior training. The user study has confirmed the accuracy.

e. Menura Supurba: Post human dreams of ersatz animals.

This paper discusses human and post-human relationships with nature and animals, using the work e. Menura Superba1 as a focal point. This interactive artwork takes the form of a Lyre bird in a cage, that mimics it’s audience in evocative ways. It is inspired by the historical practice of displaying taxidermy specimens and live species as trophies of travels to distant lands, and as symbols of wealth and status. In both form and intent the work hybridises elements from Enlightenment culture, with materials that conjure associations with dystopic post human futures (wire, post consumer electronic & other waste, as well working parts such as mobile phone screens, LED’s, camera, and cabling etc). Speculative science fiction, such as Phillip K Dick in Do Androids Dream of Electric Sheep? (Blade Runner), provides prescient stories about future (post) human worlds. This novel remains thought provoking as it describes a world that is all to rapidly approaching: where human activity has caused the destruction of most large animal species. In this fictional world, care for animals is not only a civic duty, it is one of the ways humans distinguish themselves from androids. As in Enlightenment times, ownership of animals (real, taxidermies, ersatz) is a form of commodity fetishism indicative of social status. Though whilst well heeled Victorians may have owned an elephant or have been proud of a trophy specimen, the wealthy in Dick’s future must be content with once common, even ersatz, animals such as sheep and owls, and would be repulsed to the core by the notion of killing an animal, even an ersatz animal, for sport. In becoming post human, humans have sought to separate themselves from the natural world, destroying much of it in the process. No technical prothesis will bring back to life the species we have rendered extinct. This (evolving) relationship between humanity and other species, therefore forms a central question in this work, providing a way of approaching the post human, and problematising anthropocentric perspectives. The world promised by post-human technology is indeed rich with possibility, but without corresponding steps to ensure the sustainability of technology (human society), this paper asks whether the richness of that experience will continue to be mirrored by the richness of the environments within which we exist?

A longitudinal study of health-related quality of life, fatigue and physical activity in chronic kidney disease patients

This study assessed the health-related quality of life (HRQoL), fatigue and physical activity levels of 28 persons with chronic kidney disease (CKD) on initial administration of an erythropoietin stimulating agent, and at 3 months, 6 months and 12 months. The sample comprised of 15 females and 13 males whose ages ranged from 31 to 84 years. Physical activity was measured using the Human Activity Profile (HAP): Self-care, Personal/Household work, Entertainment/Social, Independent exercise. Quality of life was measured using the SF-36 which gives scores on physical health (physical functioning, role-physical, bodily pain and general health) and mental health (vitality, social functioning, role-emotional and emotional well-being). Fatigue was measured by the Fatigue Severity Scale (FSS). Across all time points the renal sample engaged in considerably less HAP personal/household work activities and entertainment/social activities compared to healthy adults. The normative sample engaged in three times more independent/exercise activities compared to renal patients. One-way Repeated measures ANOVAs indicated a significant change over time for SF-36 scales of role physical, vitality, emotional well-being and overall mental health. There was a significant difference in fatigue levels over time [F(3,11) = 3.78, p<.05]. Fatigue was highest at baseline and lowest at 6 months. The more breathlessness the CKD patient reported, the fewer activities undertaken and the greater the reported level of fatigue. There were no significant age differences over time for fatigue or physical activity. Age differences were only found for SF-36 mental health at 3 months (t=-2.41, df=14, p<.05). Those younger than 65 years had lower emotional well-being compared to those aged over 65. Males had poorer physical health compared to females at 12 months. There were no significant gender differences on mental health at any time point. In the management of chronic kidney disease, early detection of a person’s inability to engage in routine activities due to fatigue is necessary. Early detection would enable timely interventions to optimise HRQoL and independent exercise.