Intelligent Sensors and Machine Data Processing

Detailed topics of the lectures:

1. Description of a smart system of sensors and actuators that makes decisions and interventions based on measured data, e.g. activating certain sensors and determining the frequency of measurements. Characteristics of sensor intelligence, validation, self-calibration, pre-processing, denoising, adaptivity, reconfigurability, autonomous definition of measurement strategy.
2. Sensing temperature, displacement, acceleration, touch. Integrated sensors, CMOS image sensor. Integrated sensor fabrication technology, Micro-Electro-Mechanical System (MEMS). Off-grid power issues (battery, solar, etc.).
3. Types of micromechanical integrated sensors, pressure and tactile sensors, chemical and biomedical sensors (ISFET, ChemFET), sensors using bridges and cantilevers, devices using micro-circuit boards.
4. Networked smart sensors (Radio-Frequency Identification tags for identification, etc.), their features (limited power supply, network design capability, use of machine learning methods), purpose (data collection, data analysis, decision making and data sharing).
5. Cloud computing advantages (sending large amounts of data to remote servers with unlimited capacity) and disadvantages (for time-sensitive data, the transmission time between the cloud and the intelligent sensors at the edge of the network can be long). Extending the cloud to the edge of the network: fog (Cisco)/edge (IBM)/cloudlet computing, features (distributed computing, mobility, large number of nodes, heterogeneous wireless network).
6. Evolution of sensor architectures, generations: from simple sensors to smart sensors with embedded processors; smart sensors in applications. Adding processing capabilities to devices, local processing of time-sensitive data and control of actuators.
7. Analysing data from sensors, removing outliers, source selection, handling large amounts of data, pattern recognition, classifying measured data, clustering, factor analysis, multidimensional scaling.
8. A detailed description of the different types of clustering (hierarchical or non-hierarchical; exclusionary, overlapping or fuzzy; full or partial, etc.). Types of clustering (well-differentiated, centroid-based, neighbourhood-based, density-based). Commonly used clustering methods (nearest neighbour, furthest neighbour, UPGMA, WPGMA, Ward, K-nearest neighbour, DBSCAN, etc.).
9. Advanced data mining methods to study objects that evolve over time: computing phenograms and cladograms, and applying them to the analysis of sensor data.
10. Body Worn Networks (BWN), wireless sensor networks and communication interfaces. Telemetry systems in telemedicine, application of mobile phone network based systems. Vehicular networks, Internet of Vehicles (IoV), Vehicular ad hoc network (VANET).
11. Distributed data mining: to facilitate the collection of data for spatial data mining, e.g. for monitoring air pollution. Main characteristics of this type of network (e.g. typically similar readings from nearby sensor nodes monitoring the environment). The need for data aggregation within the network due to the spatial correlation between sensor observations resulting from this type of data redundancy.
12. Intelligent cardiology sensors, pulse, blood pressure, ECG telemetry, anemometers (air velocity meters), blood oximeters, heart attack detection, cerebral vascular catastrophe prediction. Case study: analysis of medical electronic data, hardware requirements for their implementation as software.
13. Design, communication electronics and power supply issues of sensors implanted in living tissue. Data protection of telemedicine sensor networks, privacy security.
14. Review of the semester curriculum, outlook.

Detailed topics of the exercises/labs:

   1-2. Designing sensor measurements. Tasks related to feature engineering.

   3-4. Overview of possible types of data collected by sensors and related tasks.

   5-6.  Analysis of measured data using hierarchical clustering; dendrogram calculation I.

   7-8. Analysis of measured data using hierarchical clustering; dendrogram calculation II.

 9-10. Examination of measured data using non-hierarchical clustering procedures, simple practical exercises.

11-12.  Exploring relationships between data using ordination methods (factor analysis, principal component analysis, multivariate scaling) and comparing them using simple examples.