Speech detection method analysis and intelligent structure development

Making a decision about the presence/absence of the speech signal in a noisy environment is a very important problem which must be solved in many application areas including speech recognition. The paper presents a methodology analysis of several algorithms, identifying both their advantages and disadvantages to examine their ability to detect certain properties of speech. From these results, a speech detection system is designed using algorithms which complement each other. The system includes an application of 4th order cumulants and zero crossing methods to determine the presence of speech. The system is then tested up to noise levels of -4 dB while maintaining responsive detection. An adaptive fuzzy rule base is then proposed for implementation and operation under varying environmental conditions