Scribd
Upload
3 views1 page

Neural Network for Spectral Component Recognition

This document discusses using a neural network to recognize spectral components of signals. It describes training a feedforward neural network with 64 input neurons, 32 output neurons, and a hidden layer of 16 neurons to compute the spectra of signals with different frequency components and amplitudes. The network is trained and then tested on various input signals to determine if it can recognize the spectra, including noisy signals depending on the noise amplitude.

Uploaded by

Laurențiu Balaban
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
3 views1 page

Neural Network for Spectral Component Recognition

This document discusses using a neural network to recognize spectral components of signals. It describes training a feedforward neural network with 64 input neurons, 32 output neurons, and a hidden layer of 16 neurons to compute the spectra of signals with different frequency components and amplitudes. The network is trained and then tested on various input signals to determine if it can recognize the spectra, including noisy signals depending on the noise amplitude.

Uploaded by

Laurențiu Balaban
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOC, PDF, TXT or read online on Scribd

Neural recognition of spectral components of a signal Abstract

Abstract
Neural networks are used in many different application domains in order to solve various information processing problems. They have proven to be successful in pattern recognition, pattern classification, function approximation, prediction, optimization, and controlling. The neural networks are suited for solving very difficult processing problems because of some of their characteristics: the nonlinearity, the facility of learning from the environment, supervised or unsupervised and the universal approximation property. In this paper, it is approached a well known problem in signal processing, recognition of spectral components of a signal using a neural network. After studies and testing of different kind of networks it is obtained a feed forward neural network that computes the spectra of different signals of different frequency components and amplitudes. The code is very flexible and many kinds of signals can be generated for training and testing of the network. The network has 64 input neurons, 32 output neurons and one hidden layer with 16 neurons. The transfer function used in all the layers is purelin, but many other transfer functions have been tested and compared. After the training, different input signals are tested and the results are presented. The network can determine the spectra of different signals but with a step frequency given in the initial conditions. Also noisy signals are recognized but depending on the amplitude of the noise.

You might also like