Neural network signal filters are more adaptive filters than Wiener filters where resulting signal is a product of minimisation mean-square error, Wiener filter is linear filter which is adapted for some specific environment. But linear adaptive filters are limited when noise has Gaussian process pattern. So if significant Gaussian noise is present then there is need to use non-linear filters, which can deal with such noise. Also non-linear filters do not need mathematical analysis in order to calculate reference frame signal, what is not always ease to do. Non-linear adaptive filters may deal with this in two ways: improving learning efficiency and a broadening of application areas.

A neural network is a massively parallel distributed processor that has a natural propensity for storing experiential knowledge and making it available for use. It resembles the brain in two respects:

• Knowledge is acquired by the network through a learning process.

• Interconnection strengths known as synaptic weights are used to store the knowledge.

Basically, learning is a process when free parameters (i.e., synaptic weights and bias levels) of a neural network are adapted and adjusted through a continuing process of stimulation by the environment in which the network is embedded. The learning type is determined by the manner in which the parameter changes take place. Specifically, learning machines may be classified as follows:

• Learning with a teacher(also referred to as supervised learning);

• Learning without a teacher.

This second class of learning machines may also be subdivided into

• Reinforcement learning;

• Unsupervised learning or self-organized learning.

In the context of adaptive signal-processing applications, neural networks offer the following advantages:

• Non-linearity, which makes it possible to account for the non-linear behavior when generating the input data;

• The ability to approximate any prescribed input-output mapping of a continuous nature;

• Weak statistical assumptions about the environment, in which the network is embedded;

• Learning capability, which is accomplished by undertaking a training session with input-output examples that are representative of the environment;

• Generalization, the ability of the neural network to provide a satisfactory performance in response to test data never seen by the network before;

• Fault tolerance, which means that the network continues to provide an acceptable performance despite the failure of some neurons in the network

• VLSI implementability, which exploits the massive parallelism built into the design of a neural network.

This is indeed an impressive list of attributes, which accounts for the widespread interest in the use of neural networks to solve signal-processing tasks that are too difficult for linear adaptive filters.

Blogsphere: TechnoratiFeedsterBloglines
Bookmark: Del.icio.usSpurlFurlSimpyBlinkDigg
RSS feed for comments on this post  |  TrackBack URI for this post

This entry was posted on Saturday, July 14th, 2007 at 10:40 pm and is filed under Biomedical DSP. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.