About this opportunity:

Wouldn’t it be nice to toggle the lights with a simple clap, toggle music with two quick claps or control any other device?

In this Master Thesis you would investigate algorithm(s) to detect claps and to make it more interesting as well as useful, see what needs for the algorithm to work with various background noise like music or people talking. The focus should be on machine learning (ML) approaches, however a more simplistic algorithm like checking the signal amplitude could be done first for comparison purposes.

With 6G it is interesting for us to learn more about ML based receiver approaches. In Ericsson, the RF signals are sampled by an antenna and the signal is processed by a multi-core platform inside the base station. The focus in this thesis is quite similar except that the signal will be on a lower frequency so instead of an antenna we will use a microphone to capture audible sounds. Additionally, a laptop will be used to process the signal instead of a more powerful base station.

Objective:

An example approach to the work would be as follows.

• Record audio data files and label them with the type of clap in the data and other useful meta-data.
• Single clap, clap-clap and clap-clap-clap are of interest.
• Explore the data, what does a clap look like in time domain and frequency domain?
• Build a test framework that reads the files, run the algorithm on one audio-file at a time and compare with the label if the algorithm did the correct detection or not.
• Explore ML-based algorithms and benchmark them in the test framework, how many of the sound files can the algorithm pass? Start with a simple algorithm and build more advanced algorithms as time allows.

The skills you bring:

– Students of computer science, computer engineering, or similar program

– Experience in at least 1 programming language

Extent: 2 students, 30 hp each