Neural Acoustic Temperature Profiler

The purpose of Phase I of this project is to determine the feasibility of developing an acoustic instrument for measuring temperature profiles using a combination of neural networks and broadband sonar. "Broadband sonar provides the necessary comparative information between frequencies to provide a great deal of the requisite information for determining temperature from sound speed. Because the acoustic scatterers do not always obey a Raleigh distribution, which allows separation of the temperature and salinity terms for sound propagation, a neural network will be used to learn the mapping between broadband sonar echoes and the temperature in a given depth cell. During Phase I, Scientific Fishery Systems, Inc. (SFS) and RD Instruments (RDI) will combine their broadband sonar systems with additional information (geolocation, surface temperature, and date) to produce a proof-of-principle demonstation neural acoustic temperature profiler. Performance evaluation will determine the feasibility of this approach. Bridge funding will be used to design an engineering prototype of the system . Phase II will build and field test the system in a wide variety of aquatic environments to determine the limits of its capabilities. Anticipated Military Benefits: The temperature profiling system developed in this effort will find military uses in assessing ownship sound profiles and sidescan sonar mapping and minehunting. Commercial applications are tremendous in the fisheries and ocean sciences.