| Full text | |
| Author(s): |
Sammarco, Matteo
;
Zeffiro, Trevor
;
Gantert, Luana
;
Campista, Miguel Elias M.
Total Authors: 4
|
| Document type: | Journal article |
| Source: | 2024 IEEE INTERNATIONAL CONFERENCE ON PERVASIVE COMPUTING AND COMMUNICATIONS WORKSHOPS AND OTHER AFFILIATED EVENTS, PERCOM WORKSHOPS; v. N/A, p. 6-pg., 2024-01-01. |
| Abstract | |
Smart cities and Intelligent Transportation Systems rely upon the deployment of sensors in strategic areas for such purposes as crime prevention, urban planning, and road safety. In this paper, we rely on the pervasiveness of smartphones and microphones inside moving vehicles to propose a sound-based event detection system which does not require static sensing infrastructure. We train an embedded Deep Neural Network model able to identify potentially dangerous events like car accidents or emergency vehicles approaching from recorded sounds. We evaluate our model on a large novel dataset of sounds recorded inside the car cabin with audio data augmentation techniques applied thereon. We further evaluate model performance after model quantization, or the addition of environmental noise. Results show an excellent detection accuracy for dangerous events achieving a Matthews Correlation Coefficient (MCC) of 0.95. (AU) | |
| FAPESP's process: | 15/24485-9 - Future internet for smart cities |
| Grantee: | Fabio Kon |
| Support Opportunities: | Research Projects - Thematic Grants |
| FAPESP's process: | 14/50937-1 - INCT 2014: on the Internet of the Future |
| Grantee: | Fabio Kon |
| Support Opportunities: | Research Projects - Thematic Grants |