Bayesian classifier based on discretized continues feature space

摘要

The Bayesian decision theory is widely used in pattern recognition and signal detection. Only when Class-conditional-probability density is known, the theory can be used. A discretization method of stochastic variable (features) space of class-conditional-probability-density and estimation method for class-conditional-probability-distribution are proposed. Bayesian classification algorithm based on the methods is given. Finally, the methods are illustrated by applying it to recognize radar targets.