Nonlinear Noise Reduction Scheme Based on Information Flow

We present a measurement noise reduction scheme based on information flow of a chaotic system. This scheme operates on conditions of chaoticity and well-defined noise level, not depending on other detailed characteristics of noise. Starting with a simple map and full knowledge of dynamics, we extend the basic idea to general form applicable to higher dimensional systems. Reducing noise in Lorenz system is demonstrated as an example. Inferring dynamics without a priori knowledge is then discussed by proposing an indicator which measures predictability.