Course pre-requisite(s): Digital Signal Processing, Fundamentals of artificial neural networks, Python.

Course Overview

The course aims to use signal fusion in images for the purpose of using artificial intelligence techniques to perform identification or information generation tasks (classification or regression). The first part includes a review of Python and libraries such as TensorFlow (and Keras) in their application in artificial intelligence. The second part includes the generation, recording and storage of electronic sensor signals; this also includes the development of the dataset of the same. The third is the generation of images from electronic signals by positional methods, temporal characterization and through transformations. The last part is dedicated to the identification and classification of events from signals using techniques specific of artificial neural networks.

Learning Outcomes

• At the end of the course, students should be able to integrate signals to form images using different techniques.
• At the end of the course, students should be able to apply Deep Learning techniques to perform classification tasks on multi-channel information from electronic sensors.

Course Content

• Development Environment. Review of Python, Google Colab, use of Pandas, C2V, and TensorFlow libraries. Introduction to the Keras API.
• Acquisition of signals from electronic sensors and creation of Dataset. In this session, we will address the use of electronic sensors (IMU – Inertial Movement Unit, or microphones) and the acquisition, visualization and recording of the signals will be done. The topic of creating the signal dataset will also be addressed.
• Image creation. In this session, we will review some techniques to create images by integrating the sensor signals. Techniques will be used with raw signals and from obtaining temporal characteristics or through transformations.
• Identification and classification of events. From the images obtained in the previous session, solutions are developed with artificial neural networks (DNN, CNN, RNN).
• Application example. Different examples developed by master's and doctoral students from the University will be presented.

Instructional Method

• Master class. Explanation of the basic concepts related to the development environment, libraries, processing techniques, image creation and neural networks.
• Reading of technical documents related to the fundamentals of artificial neural networks and signal processing.
• Teamwork (2 students) using sensor signals provided by the teacher.

Required Course Materials

• Computer with internet connection
• Personal Google account in order to use Colab
• Personal GitHub account

Assessment

There will be four tasks to evaluate the course:
1. Team task on development environment fundamentals. Check the knowledge acquired in the management of the development environment.
2. Team task on visualization and signal processing.
3. Team task on image creation using different techniques.
4. Team task on designing a solution with neural networks for the identification and classification of events on signals.