Case Studies for Applications of Elman Recurrent Neural Networks
Elif Derya, Mustafa �beyli · InTech eBooks · 2008
ANNs may offer a potentially superior method of biomedical signal analysis to the spectral analysis methods. In contrast to the conventional spectral analysis methods, ANNs not only model the signal, but also make a decision as to the class of signal. Another advantage of ANN analysis over existing methods of biomedical signal analysis is that, after an ANN has trained satisfactorily and the values of the weights and biases have been stored, testing and subsequent implementation is rapid. The proposed combined Lyapunov exponents/RNN approach can be evaluated in discrimination of other Doppler ultrasound signals or timevarying biomedical signals. Preprocessing, feature extraction methods and ANN architectures are the main modules of an automated diagnostic systems and therefore they play important roles in determining the classification accuracies. Thus, further work can be performed for improving the classification accuracies by the usage of different preprocessing (different filtering methods), feature extraction methods (different spectral analysis methods) and ANN architectures (self-organizing map, radial basis function, mixture of experts, etc.) (?beyli, 2008a). The research demonstrated that the wavelet coefficients and the Lyapunov exponents are the features which well represent the ECG signals and the RNN trained on these features achieved high classification accuracies. The overall results of the RNN were better when they were trained on the computed composite features for each ECG beat. The results demonstrated that significant improvement can be achieved in accuracy by using the RNNs compared to the feedforward neural network models (MLPNNs). This may be attributed to several factors including the training algorithms, estimation of the network parameters and the scattered and mixed nature of the features. The results of the present study