A bezier curve approximation of the speech signal in the classification process of laryngopathies

The research concerns a computer-based clinical decision support for laryngopathies. The classification process is based on a speech signal analysis in the time domain using recurrent neural networks. In our experiments, we use the modified Elman-Jordan neural network. In the preprocessing step, an original signal is approximated using Bezier curves and next the neural network is trained. Bezier curve approximation reduces the amount of data to be learned as well as removes a noise from the original signal.