Remote Machine Learning for Cyberphysical Systems
Note that this lab is offered as a standalone course for students interested in gaining some insight into the applications of machine learning to physical systems and is not yet part of the official curriculum.
Overview
- The motivation for this lab is to help educate the students of RWTH on how to apply a selection of machine learning algorithms to real practical systems. The learning material offered here is created to allow the student to study independently and remotely without having to attend a traditional class.
- Discrete Localization: In this experiment, acoustic discrete localization is of interest, as they can achieve high accuracy, with low power consumption and infrastructure costs. The room impulse response may be used to capture a unique signature depending on the source and the receiver locations. This will be accomplished by dividing the room into smaller areas and identifying the area the new recordingis most likely to belong to. For this purpose we introduce two techniques for localization named k-nearest-neighbor and k-means clustering as well as the necessary preprocessing steps.
- Blind Source Separation: In this experiment we look at blind source separation in the context of the cocktail party problem.This describes the human ability to extract and recover a desired voice in an environment withmultiple speakers. We cover the required preprocessing steps and fundamentals of PCA and ICA and learn how to implement the these algorithm for separating speech signals.
Learning Material
- All the learning material for this lab is accessible to RWTH students via Gitlab, here.
Contact
© INDA at RWTH Aachen