Основні підходи до розпізнавання мовленнєвої інформації (частина 2)

Abstract

The paper considers and analyzes various approaches of speech signal processing and recognition. Wavelet and Fourier transform are compared as methods of initial signal processing and parameterization and artificial neural networks and artificial immune systems are compared as methods of speech recognition. ANN for speaker-dependent speech recognition is examined and ANN model for FIR and IIR digital filter simulation is presented. AIS speaker-dependent speech recognition system is investigated as the most promising among existing systems.