VEHICLE CLASSIFICATION USING NEURAL NETWORKS BASED UPON ACOUSTIC SIGNALS

This paper presents a method of vehicle classification using neural networks based upon acoustic signals detected by the AT&T SmartSonic Sensor as inputs to the networks. All data were collected from a sensor located on highway Route 460 near the Virginia Tech campus in southwest Virginia. All vehicle classifications are performed in the time domain. The feedforward multilayer perceptron trained with error backpropagation is the main algorithm used in this classification task, although a radial basis function and K nearest neighbor classifiers are used for comparison. The K nearest neighbor is the only classical method used to compare with neural network approaches. The nature of the data made decision tree algorithms ineffectual.

• Corporate Authors:

  ITS America

  1100 17th Street, NW, 12th Floor
  Washington, DC  United States  20036
• Authors:
  • James, R D
  • Sampan, S
• Conference:
  • Intelligent Transportation: Serving the User Through Deployment. Proceedings of the 1995 Annual Meeting of ITS America.
  • Location: Washington, D.C.
  • Date: 1995-3-15 to 1995-3-17
• Publication Date: 1995

Language

• English

Media Info

• Features: Figures; References;
• Pagination: p. 975-982

Subject/Index Terms

• TRT Terms: Acoustic signal processing; Networks; Sensors; Vehicle classification
• Geographic Terms: Virginia
• Old TRIS Terms: Acoustic signals
• Subject Areas: Highways; Operations and Traffic Management; Public Transportation; Vehicles and Equipment; I90: Vehicles;

Filing Info

• Accession Number: 00711375
• Record Type: Publication
• Files: TRIS
• Created Date: Sep 5 1995 12:00AM