This paper examines the benefits of edge mining -data mining that takes place on the wireless, battery-powered, and smart sensing devices that sit at the edge points of the Internet of Things. Through local data reduction and transformation, edge mining can quantifiably reduce the number of packets that must be sent, reducing energy usage, and remote storage requirements. In addition, edge mining has the potential to reduce the risk in personal privacy through embedding of information requirements at the sensing point, limiting inappropriate use. The benefits of edge mining are examined with respect to three specific algorithms: linear Spanish inquisition protocol (L-SIP), ClassAct, and bare necessities (BN), which are all instantiations of general SIP. In general, the benefits provided by edge mining are related to the predictability of data streams and availability of precise information requirements; results show that L-SIP typically reduces packet transmission by around 95% (20-fold), BN reduces packet transmission by 99.98% (5000-fold), and ClassAct reduces packet transmission by 99.6% (250-fold). Although energy reduction is not as radical because of other overheads, minimization of these overheads can lead up to a 10-fold battery life extension for L-SIP, for example. These results demonstrate the importance of edge mining to the feasibility of many IoT applications.

The fixed-size non-overlapping sliding window (FNSW) and fixed-size overlapping sliding window (FOSW) approaches are the most commonly used data-segmentation techniques in machine learning-based fall detection using accelerometer sensors. However, these techniques do not segment by fall stages (pre-impact, impact, and post-impact) and thus useful information is lost, which may reduce the detection rate of the classifier. Aligning the segment with the fall stage is difficult, as the segment size varies. We propose an event-triggered machine learning (EvenT-ML) approach that aligns each fall stage so that the characteristic features of the fall stages are more easily recognized. To evaluate our approach, two publicly accessible datasets were used. Classification and regression tree (CART), k-nearest neighbor (k-NN), logistic regression (LR), and the support vector machine (SVM) were used to train the classifiers. EvenT-ML gives classifier F-scores of 98% for a chest-worn sensor and 92% for a waist-worn sensor, and significantly reduces the computational cost compared with the FNSW- and FOSW-based approaches, with reductions of up to 8-fold and 78-fold, respectively. EvenT-ML achieves a significantly better F-score than existing fall detection approaches. These results indicate that aligning feature segments with fall stages significantly increases the detection rate and reduces the computational cost.

The high rate of falls incidence among the elderly calls for the development of reliable and robust fall detection systems. A number of such systems have been proposed, with claims of fall detection accuracy of over 90% based on accelerometers and gyroscopes. However, most such fall detection algorithms have been developed based on observational analysis of the data gathered, leading to thresholds setting for fall/non-fall situations. Whilst the fall detection accuracies reported appear to be high, there is little evidence that the threshold based methods proposed generalise well with different subjects and different data gathering strategies or experimental scenarios. Moreover, few attempts appear to have been made to validate the proposed methods in real-life scenarios or to deliver robust fall decisions in real-time. The research here uses machine learning and particularly decision trees to detect 4 types of falls (forward, backward, right and left). When applied to experimental data from 8 male subjects, the accelerometers and gyroscopes based system discriminates between activities of daily living (ADLs) and falls with a precision of 81% and recall of 92%. The performance and robustness of the method proposed has been further analysed in terms its sensitivity to subject physical profile and training set size.

The paper demonstrates that wearable sensor systems, coupled with real-time on-body processing and actuation, can enhance safety for wearers of heavy protective equipment who are subjected to harsh thermal environments by reducing risk of Uncompensable Heat Stress (UHS). The work focuses on Explosive Ordnance Disposal operatives and shows that predictions of UHS risk can be performed in real-time with sufficient accuracy for real-world use. Furthermore, it is shown that the required sensory input for such algorithms can be obtained with wearable, non-intrusive sensors. Two algorithms, one based on Bayesian nets and another on decision trees, are presented for determining the heat stress risk, considering the mean skin temperature prediction as a proxy. The algorithms are trained on empirical data and have accuracies of 92.1 ± 2.9% and 94.4 ± 2.1%, respectively when tested using leave-one-subject-out cross-validation. In applications such as Explosive Ordnance Disposal operative monitoring, such prediction algorithms can enable autonomous actuation of cooling systems and haptic alerts to minimize casualties.