Artificial Intelligence Technique for Speech Recognition based on Neural Networks

摘要

Creation of natural human sources to communicate with the computer is currently one of the greatest challenges of modern science. The speech input facility is the most user-friendly way, adopted by development of speech recognition based on sophisticated technologies. Scientists began the selection of informative signs, describing the voice signal, afterwards the task of classification of speech signals as a set of informative signs. The development of methods of signal processing in the absence of sufficient models lead to questions about the processes of generation signals using artificial neural networks, As a resu when building a signal processing system, the structure of the network should be selected; according to parameters of the signals and training network using an algorithm to maximize the use of the information contained in the data of the experiment. This article proposes the Application of wavelet transform for reduction of the value of artificial neural networks for speech recognition tasks this method a present Study a new modification of neural networks a neural network with inverse Wavelet Decomposition of the signal. For example, the speech recognition task, the analysis of the proposed method. The effectiveness of the method is proved by the results of computer simulations.