Adaptive Approach to Acoustic Car Driving Detection in Mobile Devices

P. Aksamit Opole University of Technology Faculty of Electrical Engineering, Automatic Control and Informatics, Prószkowska 76, 45-758 Opole, Poland

Full Text PDF

The context awareness of mobile devices is broadly researched area as it improves the functionality, usability, safety of usage and intelligence of the device as perceived by its user. The special type of context is driving a car. The awareness of a mobile device whose user drives a car may allow to disable some features like making and taking phone calls and enable other features like e.g. navigation thus improving the safety of the user. The paper presents the results of a research on acoustic detection of car driving based on over 60 h of collected data. The modification of traditional k-nearest neighbors classification algorithm is proposed that allows for learning and adaptation of classifier configuration. The proposed approach significantly improves both the sensitivity and specificity of the classifier comparing to the classifier based only on offline training data. The challenges in performing the acoustic wave analysis using highly heterogeneous devices like mobile phones are discussed.

DOI: 10.12693/APhysPolA.124.381 PACS numbers: 43.60.Vx, 43.60.-c, 43.60.+d, 43.60.Mn