A robust automated speech classification using hybrid wavelet-based architecture

In this paper, an efficient feature extraction approach, based on wavelet packet entropy, is proposed and used for intelligent automated speech recognition systems. A new architecture is introduced based on the integration between the wavelet packet transform and the neural network classifier model to effectively extract the features from pre-processing real English speech word signals for the purpose of automatic speech recognition among variety of speakers. The wavelet packet layer is used for optimum feature extraction in the time-frequency domain and is composed of wavelet packet decomposition and wavelet packet entropies. A multi-layer perception network is used for classification purpose. Enormous experiments have been conducted to test the efficiency of the proposed architecture. The performance of the developed system has been evaluated by using noisy English speech/voice signals. The test results showed that this system was effective in detecting real speech signals and high classification rates were obtained.