Classification of coffee aromas using e-Nose and ML/KI

Context

Coffee is a topic you wouldn't expect to hear about at ITIV, except from our lovely coffee maker Berta. On a recent project visit to Ethiopia, often referred to as the origin of coffee, we discussed the different flavors that are present in coffee. These aromas are usually analyzed by experts. To us, this seems to be a very error-prone approach and anyone who has ever bought artisanal coffee from their favorite roaster can probably confirm that not all exotic fruits can be tasted in the finished product. With new organometallic materials, aromatic compounds can now be detected with portable devices. With these electronic noses, we want to develop a system that can classify the aroma of coffee beans according to human perception.

To obtain the samples and expertise, we work together with roasting companies in Karlsruhe. During the work you will prepare samples of different coffee beans and analyze them with an electronic nose to create a digital fingerprint. The aim of the work is to investigate different machine learning approaches to link the fingerprints generated by the electronic nose with the human perception of the coffee aroma. Another approach is to investigate whether a hybrid ML model that also takes the roaster's expertise into account delivers better results.

What are you waiting for, there's free coffee after all?

Tasks

• Analysis of different coffee samples with the E-Nose
• Evaluation of different ML architectures for linking fingerprint / aroma
• Validation of the overall system in a taste seminar

Prerequisites

• Motivation and interest in solving technical problems independently
• Interest in electronics and data science
• Experience with programming (Python)
• Tea drinkers are at a disadvantage :)