论文信息 - Identification of high noise time series signals using hybrid ARMA modeling and neural network approach

Identification of high noise time series signals using hybrid ARMA modeling and neural network approach

A novel approach for time series signal identification with high noise background is proposed. The approach takes advantage of both the autoregressive moving average (ARMA) spectrum estimator and artificial neural networks (ANNs). The dynamic data system (DDS) modeling strategy and ARMA spectrum estimator are used to provide a high resolution spectrum estimate, and the backpropagation ANN is used as the feature pattern classifier. Simulation experiments based on vibration signal diagnosis are presented. The approach demonstrates a better performance than the conventional fast Fourier transform (FFT)-ANN approach in high noise environments.<<ETX>>

X. Feng | J. Schulteis | X. Feng | J. Schulteis

[1] Terrence J. Sejnowski,et al. Analysis of hidden units in a layered network trained to classify sonar targets , 1988, Neural Networks.

[2] R. Gan.,et al. An extended FFT algorithm for ARMA spectral estimation , 1984 .

[3] Shien-Ming Wu,et al. Time series and system analysis with applications , 1983 .

[4] Stephen Grossberg,et al. Studies of mind and brain , 1982 .

[5] Jerry M. Mendel,et al. Use Of Higher-Order Statistics In Signal Processing And System Theory: An Update , 1988, Optics & Photonics.

[6] R. A. Clinkenbeard,et al. An unsupervised learning and fuzzy logic approach for software category identification and capacity planning , 1992, [Proceedings 1992] IJCNN International Joint Conference on Neural Networks.