Bio-inspired method to distinguish
man-made objects in the ocean.
Shaun D Anderson
Georgia Institute of Technology
Woodruff School of Mechanical Engineering
Graduate Box 1000
Atlanta, Ga 30332-0405, USA
Karim G. Sabra
Georgia Institute of Technology
Woodruff School of Mechanical Engineering
771 Ferst Drive NW
Atlanta, Ga 30332-0405, USA
Manell E. Zakharia
French Naval Academy
BP 600 29240 Brest-Armees, France
Mario Zampolli
NATO Undersea Research Centre
Viale San Bartolmeo 400
19126 La Spezia, Italy
Henrik Schmidt
Massechusest Institute of Technology
77 Mass Ave, 5-204
Cambridge, MA 02139, USA
William A. Kuperman
MPL, Scripps Institute of Oceanography
University of California, San Diego
La Jolla, CA 92093-0238, USA
Popular version of paper 5pUWb8
"Robust recognition and characterization of man-made objects in shallow water using time-frequency analysis”
A familiar “ding ding ding” is heard over the conversations in a room as someone prepares for a toast. From this familiar event, a trained ear would be able to use this ringing sound to distinguish if that glass were made from crystal. The unique ring created by tapping a glass is known as a resonance frequency. This seemingly simple process is truly an amazing and tremendously complex tasks the human ear is capable of performing. This single event requires a person to separate certain tones, decipher times at which they occur, and then search memory to distinguish the material of the glass. Additional complexity is added to this process due to ambient noise in the room distorting the sound actually heard.
Currently there is a desire to be able to find and classify man-made objects in the ocean with out having to waste time analyzing rocks, or scanning every inch of the ocean bottom. This is possible by using sound, which travels extremely well in the ocean. The difficulty of this method becomes how to process sound recorded in the ocean in-order to identify an object of interest. The bio-inspired method of separating time and frequency is ideal for this problem.
Being able to separate and identify a resonance frequency like the human ear becomes extremely useful for this classification problem. Classification is possible due to distinct “pitch” or resonance of a man-made object caused by the methodical design. Alternatively natural objects such as rocks do not have as distinct of a resonance caused by the randomness of natural structures. Additional benefits of this method emerge because noise is typically broadband which allows the resonance to be enhanced above the background noise.
A robust method implementing the time-frequency separation for classification has been developed. By exciting objects on the sea-floor with a sound source and recording the response, allow the resonance of an object to be identified in a noisy environment. Recorded sound data is then processed with the time-frequency analysis to separate ringing and classify any objects. Additional information such as material, size, and shape of an object can be extracted from the recorded sound. This process is made much more robust by the use of time-frequency analysis.
In-order to extract this additional information for a large variety of objects of interest in the ocean, a model was created to “teach” the computer different resonance signatures for objects of interest such as sea mines. Additionally the environment in which the object is located becomes important to the response of the object, and it too must be added to the model.
Thus the method mimics the complexity of the human processing of sound to identify resonance, such as the ring of a crystal glass or a musician tuning an instrument by ear. The benefits of the time-frequency processing technique lends itself perfectly to object classification in the ocean due to resonance isolation and noise reduction. By identifying an objects resonance information about the objects size, material, and structure can be determined.