The use of electroencephalography (EEG) has recently grown as a means to diagnose neurodegenerative pathologies such as Alzheimer’s disease (AD). AD recognition can benefit from machine learning methods that, compared with traditional manual diagnosis methods, have higher reliability and improved recognition accuracy, being able to manage large amounts of data. Nevertheless, machine learning methods may exhibit lower accuracies when faced with incomplete, corrupted, or otherwise missing data, so it is important do develop robust pre-processing techniques do deal with incomplete data. The aim of this paper is to develop an automatic classification method that can still work well with EEG data affected by artifacts, as can arise during the collection with, e.g., a wireless system that can lose packets. We show that a recurrent neural network (RNN) can operate successfully even in the case of significantly corrupted data, when it is pre-filtered by the robust principal component analysis (RPCA) algorithm. RPCA was selected because of its stated ability to remove outliers from the signal. To demonstrate this idea, we first develop an RNN which operates on EEG data, properly processed through traditional PCA; then, we use corrupted data as input and process them with RPCA to filter outlier components, showing that even with data corruption causing up to 20% erasures, the RPCA was able to increase the detection accuracy by about 5% with respect to the baseline PCA.

The human activity monitoring technology is one of the most important technologies for ambient assisted living, surveillance-based security, sport and fitness activities, healthcare of elderly people. The activity monitoring is performed in two steps: the acquisition of body signals and the classification of activities being performed. This paper presents a low-cost wearable wireless system specifically designed to acquire surface electromyography (sEMG) and accelerometer signals for monitoring the human activity when performing sport and fitness activities, as well as in healthcare applications. The proposed system consists of several ultralight wireless sensing nodes that are able to acquire, process and efficiently transmit the motion-related (biological and accelerometer) body signals to one or more base stations through a 2.4 GHz radio link using an ad-hoc communication protocol designed on top of the IEEE 802.15.4 physical layer. A user interface software for viewing, recording, and analysing the data was implemented on a control personal computer that is connected through a USB link to the base stations. To demonstrate the capability of the system of detecting the user’s activity, data recorded from a few subjects were used to train and test an automatic classifier for recognizing the type of exercise being performed. The system was tested on four different exercises performed by three people, the automatic classifier achieved an overall accuracy of 85.7% combining the features extracted from acceleration and sEMG signals. A low cost wireless system for the acquisition of sEMG and accelerometer signals has been presented for healthcare and fitness applications. The system consists of wearable sensing nodes that wirelessly transmit the biological and accelerometer signals to one or more base stations. The signals so acquired will be combined and processed in order to detect, monitor and recognize human activities.

Photoplethysmography (PPG) is a common and practical technique to detect human activity and other physiological parameters and is commonly implemented in wearable devices. However, the PPG signal is often severely corrupted by motion artifacts. The aim of this paper is to address the human activity recognition (HAR) task directly on the device, implementing a recurrent neural network (RNN) in a low cost, low power microcontroller, ensuring the required performance in terms of accuracy and low complexity. To reach this goal, (i) we first develop an RNN, which integrates PPG and tri-axial accelerometer data, where these data can be used to compensate motion artifacts in PPG in order to accurately detect human activity; (ii) then, we port the RNN to an embedded device, Cloud-JAM L4, based on an STM32 microcontroller, optimizing it to maintain an accuracy of over 95% while requiring modest computational power and memory resources. The experimental results show that such a system can be effectively implemented on a constrained-resource system, allowing the design of a fully autonomous wearable embedded system for human activity recognition and logging.

Falls and their aftermath pose significant healthcare challenges, impacting individuals across various age groups and occupational backgrounds. These incidents detrimentally affect functional mobility and overall quality of life, necessitating a comprehensive approach to fall detection systems in diverse populations. Therefore, wearable devices are necessary to continuously monitor activities. This work introduces a novel deep-learning model specifically optimized for edge devices capable of detecting falls. The wearable sensor integrates a pressure sensor, a three-axis gyroscope, and a three-axis accelerometer. The developed system works in real-time with the dual objective of identifying the activities carried out and classifying them as falls or daily life activities. We evaluated this approach using both our self-collected dataset and a publicly available one (SisFall). Furthermore, in our dataset, we also introduced the syncope between falls that the sensor must be able to detect. Results demonstrate that while maintaining low-cost, low-complexity of the model, low-power consumption, and high-speed data processing, combining usage of the three sensors and deep learning algorithm allows to obtain an accuracy of 99.38%, and inference time of 25 ms.

Esca is one of the most common disease that can severely damage grapevine. This disease, if not properly treated in time, is the cause of vegetative stress or death of the attacked plant, with the consequence of losses in production as well as a rising risk of propagation to the closer grapevines. Nowadays, the detection of Esca is carried out manually through visual surveys usually done by agronomists, requiring enormous amount of time. Recently, image processing, computer vision and machine learning methods have been widely adopted for plant diseases classification. These methods can minimize the time spent for anomaly detection ensuring an early detection of Esca disease in grapevine plants that helps in preventing it to spread in the vineyards and in minimizing the financial loss to the wine producers. In this article, an image dataset of grapevine leaves is presented. The dataset holds grapevine leaves images belonging to two classes: unhealthy leaves acquired from plants affected by Esca disease and healthy leaves. The data presented has been collected to be used in a research project jointly developed by the Department of Information Engineering, Polytechnic University of Marche, Ancona, Italy and the STMicroelectronics, Italy, under the cooperation of the Umani Ronchi SPA winery, Osimo, Ancona, Marche, Italy. The dataset could be helpful to researchers who use machine learning and computer vision algorithms to develop applications that help agronomists in early detection of grapevine plant diseases. The dataset is freely available at http://dx.doi.org/10.17632/89cnxc58kj.1

Vehicle and pedestrian detection (VaPD) is one of the most critical tasks in an advanced driver assistance system which help the driver to drive safely and save the pedestrian life. VaPD is a typical object detection problem that requires a trade-off among accuracy, speed, and memory consumption. Most existing methods focus on improving detection accuracy, while ignoring VaPD requires real-time detection speed with limited computational resources. Thus, it is of primary importance to study light-weight and real-time VaPD methods for embedded devices, that is hardware platforms with limited computation and memory resources. To deal with these issues, this paper proposes a low-rank (LR) Tiny YOLO v3 architecture that meets the requirements of real-time VaPD on embedded systems. The architecture has been developed starting from Tiny YOLO v3 adopting a convolutional neural network compression technique based on Tucker tensor decomposition, able to reduce the computational complexity of the network. A wide experimentation has been carried out on two embedded platforms, Raspberry Pi 4 and NVIDIA Jetson Nano 2 GB, and two datasets commonly used for VaPD, PASCAL VOC and KITTI dataset, showing the superiority of the LR Tiny YOLO v3 with respect to the state-of-the-art networks in obtaining the best compromise between inference time, accuracy and memory occupancy. Moreover, the proposed architecture meets the requirements of VaPD on embedded systems using only 22% of the memory required by the baseline Tiny YOLO v3 Darknet, and always providing better inference time (36.46 FPS) with only a marginal decrease in accuracy ( $\sim$ 2%).