The most common mode of communication between humans is speech. As this is the most preferred way, humans would like to use speech to interact with machines also. That is why; speech recognition has gained a lot of popularity. Many approaches for speech recognition exist like Dynamic Time Warping (DTW) and Hidden Markov Model (HMM). This paper is defined a three stage neural integrated model speech signal enhancement and use the decomposition integrated HMM model for speech feature transformation. For the feature extraction of speech Discrete Wavelength Transform (DWT) has been used which gives a set of feature vectors of speech waveform.