Projection Pursuit Learning Networks Applying for Eliminating Noises of Fiber Optic Gyroscope

To cancel nonlinear noise of fiber optical gyroscope's output signal.A projection pursuit learning network was developed to approach the Volterra filter.The PPLN uses the batch learning technique and parameter optimized alternately to acquire adapting networks size,weights and hidden unit functions,hence can obtain simpler architecture and good robustness.It was to overcom problem that number of the Volterra filter coefficients assumes a geometric series increase with the increasing of the filter order,it is very difficult to realize the filter.Simulation and experiment show that PPLN filter has a better result than the Volterra filter in noise cancelations.