Daniele Toscani

An integrated communications framework for context aware continuous monitoring with body sensor networks

This paper deals with a wireless pervasive communication system to support advanced healthcare applications. The proposed system is based on an ad hoc interaction of mobile body sensor networks with independent wireless sensor networks already deployed within the environments in order to allow a continuous and context aware health monitoring for patients along their daily life scenarios with an unprecedented precision and flexibility of sensing. After an accurate protocol characterization, simulation results are provided, underlining remarkable performance with respect to existing solutions, for different mobility models and node density values.

IKNOS: inference and knowledge in networks of sensors

This paper presents a framework for managing data from sensor of poor quality, with the objective to reduce at the same time the communication load and hence energy consumption. Each node in a wireless sensor network maintains a simple local model of the data it is collecting and sends its parameters to a central location (sink), where it is executed the global monitoring. Local models are used to simulate sensors readings, minimising the need of communication with sensors and hence the consumption of their battery; they are updated locally, when sensor readings differ excessively from simulated data. At the sink the global model (a Bayesian Network) is learnt on the simulated data. It is used to identify and replace anomalous readings (outliers) that a sensor should have produced and to detect anomalies missed by any single node (when communication with a sensor is interrupted).

A DSS for assessing the impact of environmental quality on emergency hospital admissions

In this paper we present a Decision Support System (DSS) aimed at forecasting high demand of admission on health care structures due to environmental pollution. The computational engine of the DSS is based on Autoregressive Hidden Markov Models (AHMM). We estimate the forecasting model by analyzing the historical daily average concentrations of pollutants and the number of hospital admissions, collected from multiple data sources. Given the actual concentration of different pollutants, measured by a network of sensors, the DSS allows to forecast the demand of hospital admissions for acute diseases in the following 1 to 6 days. We tested our system on cardiovascular and respiratory diseases in the area of Milan. The performances of our system, compared with multiple linear regression, show that AHMM are a robust approach to capture the connections among health and environmental indicators.

Classifying and counting vehicles in traffic control applications

This paper presents a machine learning system to handle traffic control applications. The input of the system is a set of image sequences coming from a fixed camera. The system can be divided into two main subsystems: the first one, based on Artificial Neural Networks classifies the typology of vehicles moving within a limited image area for each frame of the sequence; the second one, based on Genetic Algorithms, takes as input the frame-by-frame classifications and reconstructs the global traffic scenario by counting the number of vehicles of each typology. This task is particularly hard when the frame rate is low. The results obtained by our system are reliable even for very low frame rate (i.e. four frames per second). Our system is currently used by a company for real-time traffic control.