A Modern Approach to Convert Nam and Stuttered Speech into Normal Speech by Using DSP with ANN

Abstract

NAM and Stammering speech is considered as a Non audible murmuring voice which is considered as an unrecognizable speech signal. Due to the above factor it is necessary to convert the NAM and stuttered speech into recognizable speech as well as text. In order to achieve this conversion some specific procedures, techniques, algorithms are necessarily to be used in an appropriate manner. Here the conversion process are considered into four stages such that each stage perform the separate tasks in a regular time interval. In the first stage either NAM or stuttered speech signal is transferred into the recognized word by using NAM Microphone along with HMM and DWT methods. Further to this the recognized speech is modified for the speech newlinesynthesis processing whereas the parametric measurements are updated in this section and also the causes and minimization of interferences are completely analyzed such that measurement of the vibration of voice signals are also effectively carried out. The next stage is the suppression of noise such that the noise is completely eliminated by using EKF such that the nonlinear parametric estimation are effectively carried out at a regular time intervals. Here the speech signals before and after noise are compared through appropriate sound measurement techniques. Finally the noise free signals are sent into the CNN such that the error estimation levels are minimized through the CNN. It is to be noted that the different performance parameters are compared such that these are enhanced for graphical representation. The performance measures are neural network model, Parameter analysis, signal strength analysis and classified stuttered word. Here the MSE, MAE, correlation, SNR, PSNR, Abs value of SNR and PSNR are considered. newline newline