Bayesian inference in a distributed associative neural network for adaptive signal processing

The primary advantages of high performance associative memory model are its ability to learn fast, store correctly, retrieve information similar to the human "content addressable" memory and it can approximate a wide variety of non-linear functions. Based on a distributed associative neural network, a Bayesian inference probabilistic neural network is designed implementing the learning algorithm and the underlying basic mathematical idea for the adaptive noise cancellation. Simulation results using speech corrupted with low signal to noise ratio in telecommunication environment shows great signal enhancement. A system based on the described method can store words and phrases spoken by the respect to noise, regardless of its origin and level. New words, pronunciations, and languages can be introduced to the system in an incremental, adaptive mode.