Spectral estimation has been applied broadly in the signal-processing domain. In modern spectral estimation methods, the model parameters will be obtained by solving the Yule-Walker equations. There are several common disadvantages such as complicated processing steps and heavy calculation load in all the optimized algorithms. In this paper, two methods using neural networks are discussed. In the first one, the solving of the equations set is avoided. In the second one, the solving of the equations set is simplified by using a simple BP neural network. Simulations show that the spectral analysis effect is better than some other methods.
Published in:
Neural Information Processing, 2002. ICONIP '02. Proceedings of the 9th International Conference on
(Volume:5
)
Date of Conference: 18-22 Nov. 2002