Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

Human activity recognition (HAR) tasks have traditionally been solved using engineered features obtained by heuristic processes. Current research suggests that deep convolutional neural networks are suited to automate feature extraction from raw sensor inputs. However, human activities are made of complex sequences of motor movements, and capturing this temporal dynamics is fundamental for successful HAR. Based on the recent success of recurrent neural networks for time series domains, we propose a generic deep framework for activity recognition based on convolutional and LSTM recurrent units, which: (i) is suitable for multimodal wearable sensors; (ii) can perform sensor fusion naturally; (iii) does not require expert knowledge in designing features; and (iv) explicitly models the temporal dynamics of feature activations. We evaluate our framework on two datasets, one of which has been used in a public activity recognition challenge. Our results show that our framework outperforms competing deep non-recurrent networks on the challenge dataset by 4% on average; outperforming some of the previous reported results by up to 9%. Our results show that the framework can be applied to homogeneous sensor modalities, but can also fuse multimodal sensors to improve performance. We characterise key architectural hyperparameters’ influence on performance to provide insights about their optimisation.

Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition

Human physical activity recognition based on wearable sensors has applications relevant to our daily life such as healthcare. How to achieve high recognition accuracy with low computational cost is an important issue in the ubiquitous computing. Rather than exploring handcrafted features from time-series sensor signals, we assemble signal sequences of accelerometers and gyroscopes into a novel activity image, which enables Deep Convolutional Neural Networks (DCNN) to automatically learn the optimal features from the activity image for the activity recognition task. Our proposed approach is evaluated on three public datasets and it outperforms state-of-the-arts in terms of recognition accuracy and computational cost.

Human Activity Recognition Using Wearable Sensors by Deep Convolutional Neural Networks

As smartphone penetration saturates, we are witnessing a new trend in personal mobile devices—wearable mobile devices or simply wearables as it is often called. Wearables come in many different forms and flavors targeting different accessories and clothing that people wear. Although small in size, they are often expected to continuously sense, collect, and upload various physiological data to improve quality of life. These requirements put significant demand on improving communication security and reducing power consumption of the system, fueling new research in these areas. In this paper, we first provide a comprehensive survey and classification of commercially available wearables and research prototypes. We then examine the communication security issues facing the popular wearables followed by a survey of solutions studied in the literature. We also categorize and explain the techniques for improving the power efficiency of wearables. Next, we survey the research literature in wearable computing. We conclude with future directions in wearable market and research.

A Survey of Wearable Devices and Challenges

This paper proposes a novel participant selection framework, named CrowdRecruiter, for mobile crowdsensing. CrowdRecruiter operates on top of energy-efficient Piggyback Crowdsensing (PCS) task model and minimizes incentive payments by selecting a small number of participants while still satisfying probabilistic coverage constraint. In order to achieve the objective when piggybacking crowdsensing tasks with phone calls, CrowdRecruiter first predicts the call and coverage probability of each mobile user based on historical records. It then efficiently computes the joint coverage probability of multiple users as a combined set and selects the near-minimal set of participants, which meets coverage ratio requirement in each sensing cycle of the PCS task. We evaluated CrowdRecruiter extensively using a large-scale real-world dataset and the results show that the proposed solution significantly outperforms three baseline algorithms by selecting 10.0% -- 73.5% fewer participants on average under the same probabilistic coverage constraint.

CrowdRecruiter: selecting participants for piggyback crowdsensing under probabilistic coverage constraint

Energy storage is quickly becoming the limiting factor in mobile pervasive technology. For intelligent wearable applications to be practical, methods for low power activity recognition must be embedded in mobile devices. We present a novel method for activity recognition which leverages the predictability of human behavior to conserve energy. The novel algorithm accomplishes this by quantifying activity-sensor dependencies, and using prediction methods to identify likely future activities. Sensors are then identified which can be temporarily turned off at little or no recognition cost. The approach is implemented and simulated using an activity recognition data set, revealing that large savings in energy are possible at very low cost (e.g. 84% energy savings for a loss of 1.2 pp in recognition).

Energy-Efficient Activity Recognition Using Prediction

Energy storage is quickly becoming the limiting factor in mobile pervasive technology. We introduce a novel method for activity recognition which leverages the predictability of human behavior to conserve energy by dynamically selecting sensors. We further present a taxonomy of existing approaches to dynamically reducing consumption while maintaining recognition rates. The novel algorithm conserves energy by quantifying activity-sensor dependencies and using prediction methods to identify likely future activities. The approach is implemented and simulated using two activity recognition data sets, and the effects of the novel method are evaluated in terms of recognition rates, energy consumption, and prediction rates. The results indicate that switching off sensors only significantly affects prediction under extreme conditions and that these effects can be counteracted by adjusting system parameters. Large savings in energy can be achieved at very low cost, for example, recognition losses of 1.5 pp with 84.8 % energy savings for the first data set, and 2.8 pp and 89.9 % for the second.

Jurgen Czerny

Papers

Energy-Efficient Activity Recognition Using Prediction

Activity recognition for creatures of habit