During childhood development, a person learns to recognize sounds from the environment and predict what these sounds mean. The question is how to achieve this with robots that also learn to predict what certain sound characteristics mean based on previous experiences.
The master's thesis is based on the "Boombox" project. The project presents an algorithm that reconstructs the shape and location of the object based on the sound of the object falling into the box. The task involves the design of an experimental sound measurement and recording system to be used for machine learning and the design of a neural network to reconstruct the shape and location of a dropped object into a box.
Good knowledge of the Python programming language and a willingness to learn about machine learning with neural networks are required.
Contact: Janko Slavič