Available online at: www.ijarcsse.com SUV Detection Algorithm for Speech Signals

Abstract — One of the most influential device in human lives is the `Computer`. The usage of computers is increasing exponentially in every other field existing on this earth. No doubt it simplifies the way we handle data. It also helps in easily exchanging information among humans, with everyday new technologies coming up for making the sharing of data more and more easy. But it does require some input when an interface is utilized to transfer data. Currently most common form of input provided is touch or by mouse and keyboard. With pioneering research in `Speech Recognition ` from many years, it has become one of the most prominent part of modern technological systems. Thus, `Speech ` is also considered as an input to computer, making it a natural and faster way of providing input to computers. A high degree of accuracy is required for good speech recognition system. The predominant problems faced are silence detection and removal, voiced unvoiced distinguishing, detection of word boundary, noise removal. This paper discusses the implementation of an algorithm which automatically detects the silence, voiced and unvoiced parts of a speech signal, which can drastically improve the accuracy of a speech recognition system. The algorithm is based on three important characteristics of a speech Signal – Zero Crossing Rate, Short Time Energy and Fundamental Frequency