Human Activity Recognition (HAR) is currently recognized as a key element of a more general framework designed to perform continuous monitoring of human behaviors in the area of Ambient Assisted Living (AAL), well-being management, medical diagnosis, elderly care, rehabilitation, entertainment and surveillance in smart home environments. In this paper, an innovative HAR system, exploiting the potential of wearable devices integrated with the skills of deep learning techniques, is presented with the aim of recognizing the most common daily activities of a person at home. The designed wearable sensor embeds an Inertial Measurement Unit (IMU) and a Wi-Fi section to send data on a cloud service and to allow direct connection to the Internet through a common home router so that the user themselves could manage the installation procedure. The sensor is coupled to a CNN network designed to make inferences with the minimum possible resources to keep open the way of its implementation on low-cost or embedded devices. The system is conceived for daily activity monitor and nine different activities can be highlighted with an accuracy of 97%.

Location-based services have increased in popularity in recent years and can be fruitfully exploited in the field of smart homes, opening the doors to a wide range of personalized services. In this context, radio technology can be widely employed since, other than connecting devices in the home system, it offers solutions for the user localization issue without the need of any extra device. Techniques based on received signal strength indicator (RSSI) are often used, relying on fingerprinting or proximity algorithms. In this paper, a novel RSSI-based fingerprinting approach for room-level localization is presented: it is a threshold algorithm based on receiver operating characteristic analysis. Moreover, the actual user location is estimated from his/her interaction with the home system devices deployed in the house: if the home environment is inhabited by more than one person, it becomes of utmost importance the identification of who is actually interacting with a given device. A proximity method is exploited for this purpose. Tests have been carried out to characterize the approach, particularly, the effects of RSSI samples, number and position, of the anchor nodes have been analyzed. Finally, some considerations about power consumption of the mobile node have been presented.

We evaluated the effectiveness in classifying food images of a deep-learning approach based on the specifications of Google's image recognition architecture Inception. The architecture is a deep convolutional neural network (DCNN) having a depth of 54 layers. In this study, we fine-tuned this architecture for classifying food images from three well-known food image datasets: ETH Food-101, UEC FOOD 100, and UEC FOOD 256. On these datasets we achieved, respectively, 88.28%, 81.45%, and 76.17% as top-1 accuracy and 96.88%, 97.27%, and 92.58% as top-5 accuracy. To the best of our knowledge, these results significantly improve the best published results obtained on the same datasets, while requiring less computation power, since the number of parameters and the computational complexity are much smaller than the competitors?. Because of this, even if it is still rather large, the deep network based on this architecture appears to be at least closer to the requirements for mobile systems.

The activation energy of electromigration in 'bamboo'-type, passivated Al-1% Si/TiN/Ti lines was determined by means of high resolution resistance measurements at wafer level. Both the very early stages, with a nonlinear behaviour of the resistance, and the subsequent stages, characterized by an approximately constant rate of resistance change, were analysed with an existing model that correlates electromigration and mechanical stress evolution. An activation energy could be extracted only in the phase of linear resistance increase, where a value of 0.95 eV was found. Diffusion at the interface between Al-Si and the barrier metal or the passivation could be responsible for this value of the activation energy.

Food intake and eating habits have a significant impact on people’s health. Widespread diseases, such as diabetes and obesity, are directly related to eating habits. Therefore, monitoring diet can be a substantial base for developing methods and services to promote healthy lifestyle and improve personal and national health economy. Studies have demonstrated that manual reporting of food intake is inaccurate and often impractical. Thus, several methods have been proposed to automate the process. This article reviews the most relevant and recent researches on automatic diet monitoring, discussing their strengths and weaknesses. In particular, the article reviews two approaches to this problem, accounting for most of the work in the area. The first approach is based on image analysis and aims at extracting information about food content automatically from food images. The second one relies on wearable sensors and has the detection of eating behaviours as its main goal.

Brain–computer interface (BCI) systems aim to enable interaction with other people and the environment without muscular activation by the exploitation of changes in brain signals due to the execution of cognitive tasks. In this context, the visual P300 potential appears suited to control smart homes through BCI spellers. The aim of this work is to evaluate whether the widely used character-speller is more sustainable than an icon-based one, designed to operate smart home environment or to communicate moods and needs. Nine subjects with neurodegenerative diseases and no BCI experience used both speller types in a real smart home environment. User experience during BCI tasks was evaluated recording concurrent physiological signals. Usability was assessed for each speller type immediately after use. Classification accuracy was lower for the icon-speller, which was also more attention demanding. However, in subjective evaluations, the effect of a real feedback partially counterbalanced the difficulty in BCI use. Practitioner Summary: Since inclusive BCIs require to consider interface sustainability, we evaluated different ergonomic aspects of the interaction of disabled users with a character-speller (goal: word spelling) and an icon-speller (goal: operating a real smart home). We found the first one as more sustainable in terms of accuracy and cognitive effort.