Thesis

Back to list

Classification of Car Classes Based on Combustion Engine Sound

Authors Seo, J.
Year 2017
Thesis Type Audio Engineering project
Topic Audio Signal Processing
Keywords Psychoakustik, Signalverarbeitung
Abstract Abstract Since the launch of electric vehicles, the sound design for those silent vehicles has become a highly important issue nowadays, as it is required to have a suitable exterior sound to alert pedestrians to approaching vehicles as well as an interior one for driver’s recognition of the driving situation. One widely applicable method for the engine sound design is to utilise the knowl- edge of conventional combustion engine sounds. A engine of a car also generates an appropriate sound corresponding to its specific car class. Is it therefore possible to distinguish different car classes by their engine sound characteristics? The aim of this work is to answer this question, i.e. to classify the car classes based on the engine sounds specified by relevant psycho-acoustic parameters. After pre-processing the given data, including normalisation and unnecessary param- eter removal, the parameters with similar behaviour are grouped by means of PCA. The feature subset is then generated in such a way that one parameter is selected in each group, and then the classifier is constructed after applying LDA. The classifiers that distinguish specific classes are then evaluated according to their performance and robustness. Lastly, potential future works are proposed based on limitations of this work.
URL http://phaidra.kug.ac.at/o:61096
Supervisors Sontacchi, A.