Sensors for continuous glucose monitoring provide real-time data about blood glucose concentration values. The operational duration of these devices ranges between 10 to 15 days, during which they frequently exhibit known errors as documented within the user application notes. This study introduces a novel approach, employing a comprehensive synthetic dataset that simulates 500 responses of 10 CGM sensors, to facilitate their self-calibration and self-learning autonomously, bypassing the need for back-propagation. The proposed solution minimizes sensor errors utilizing a Tiny Radial Basis Function Neural Network (TinyRBF) as its foundation. A variety of calibration intervals, spanning from 1 hour to a full week, were employed to characterize the model. The TinyRBF model demonstrated a reduced demand for computational resources by employing an average of 1.02 neurons, which is less than that required by other compact models such as Legendre Memory Unit (LMU) and Temporal Convolutional Network (TCN) models. This approach yielded a Mean Absolute Error (MAE) of 12.1 mg/dL, with recalibration executed every three days. Ultimately, the TinyRBF model was implemented on an Intelligent Sensor Processing Unit (ISPU), incorporating a low-power instruction set directly within the sensor package. The model quantized to 16 bits exhibited a 33.5% decrease in inference time in contrast to its floating-point counterpart. These findings imply its suitability for implementation within the sensor’s embedded computational resources.

The identification of floating particles in liquids in order to characterize their purity and quality is a topic of growing interest in the face of the increasing attention being paid to product quality control and the rising tide of pollution in primary goods such as the drinking water. The problem of microplastics spread in water and food is one of the main issues of attention today, mainly because of the effects on people's health who consume these goods. The monitoring of large volumes of water represents one of the main issues of interest that is driving the development of non-invasive and non-destructive high-precision techniques. Among the most interesting methods of performing this monitoring, optical systems represent a solution of great interest given their negligible, if any, impact on the monitored products and their ability to continuously analyzing the compound of interest. Given a high-quality optical recording system, it is necessary to complement it with a highly reliable and fast detection system to allow large volumes to be monitored in a relatively short time. In this scenario, the current paper brings three main contributions: (i) it defines and models a detection system with controllable reliability, (ii) it presents an online detection algorithm and (iii) it tests the suitability of the proposed system for integration into existing monitoring devices.

IoT sensors networks are often characterized by stringent power requirements and a high probability of sensor fault. This paper, thanks to Graph Signal Processing (GSP), aims to model an IoT scenario to find an optimal sensor configuration for battery-saving applications. In detail, the Girwan Newman method is applied to the graph to find clusters and the performance of the described method is evaluated in terms of signal-noise ratio depending on the fraction of sampled sensors. Tests were performed both on simulated data and real data from the European Environment Agency considering several air pollutants concentrations.