Abstract: Speech is one of the important requirements and the well-situated method of communication between people. Real time speech to textis an accurate conversion of uttered words after speaking.STT is very useful tool to interact with people in counseling interviews or conference.Thus the conversion of speech to written language must be accurate and fast so that it can be easily understand by people.The fundamental approach of this paper is to develop an algorithm to convert speech to text using Punjabi phonetics. This paper introduces and discusses two popular and different noise reduction techniques (Auto Spectral Subtraction, LBG, MFCC)andpresents our simulation result of a noise reduction system. It isshown that the systemreduces the noise almost completely by finding the minimum Euclidean distance andkeeps the enhanced speech signal very similar to the originalspeech signal. This paper presents a method to design a speech to text conversion module using JAVA. This method is simple to use and takes less use of memory space.