Vehicle IdentificationDetect vehicles in large open fields

and to track the movement of vehicles remotely

Presented by

Hoda El-Dawy Mohamed

The Main Idea

• Identify vehicles using all spectrum distribution of the acoustic sound* which is unique to the vehicle.

• Sound data are captured and divided into multiple sound frames with n frequency samples per frame

• A signature is defined for each vehicle.

• Distribute the database and comparison process among N sensor nodes. Each sensor has number of reference vehicle signature where each signature has only 1/N portion of the reference vehicle signature

*Acoustics is the interdisciplinary science that deals with the study of all mechanical waves in gases, liquids, and solids including vibration ,sound, ultrasound and infrasound. When sound enters a new medium, it is reflected, transmitted, or absorbed. Sound can be absorbed by certain materials: rubber, cork, and acoustic tiles, for example. Sound absorbing materials have high absorption coefficients. The absorption coefficient is the fraction of the original sound intensity absorbed by the material at some given frequency.

The Main Achievements

•The proposed algorithm can distinguish and identify vehicles with

great accuracy as it uses the overall shape of the sound spectrum

rather than the limited number of feature vectors.

•The sensor node has limited capacity (1/N portion of the vehicle

signature).

The Challenges

•Porting the algorithm makes vehicle identification possible in wide

open areas but it is quite challenging due to the limitations of a

sensor node.

•Experiment challenges

–The microphone head is covered with cotton balls to reduce wind noise

when vehicles approach to the microphone. In addition, the microphone

sits on a sponge to block ground vibration generated by moving vehicles

and to capture the sound with less noise.

Accuracy in comparing signatures