While most electric vehicles today provide an internal driving sound, these sounds are the result of a static approach, in terms of being reproducible and only based on a limited number of vehicle-related input parameters. Notwithstanding the possibility of tuning these sound generators by driving parameters (e.g., speed and load), static sound design has the disadvantage of not being adaptable to distinct contexts, environments and personal preferences. In this paper, we present a new approach that combines machine learning and adaptive sound generators to match the detected situational driving style. The proposed solution contains two main components:(a) Morphable sound generators with hyperparameters based on the knowledge of sound design experts; (b) A deep learning model capable of predicting personal driving styles. The AI detected driving style is applied to control a variety of seamlessly morphable sound algorithms, thus providing the driver with a sonic experience dynamically adapted to her/his traf?c environment or driving style. The driver is furthermore enabled to intentionally determine the sonic behaviour of the vehicle without the use of any additional controls. A prototype of this system has been implemented and tested with several participants, providing resources for quantitative analysis of multi-modal feature selection and respective model ?tting. This was followed by interviews with participants regarding the qualitative (sound design) results of this ?rst use case.
(See document for exact affiliation information.)
https://aes2.org/publications/elibrary-page/?id=21808
(3000KB)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
