Speech Recognition Using Backoff N-Gram Modelling in Android Application

Google is one of the most popular information retrieval systems among users. Spoken questions are a natural standard for penetrating the network in settings where typing on a console is not applicable. This paper describes a speech boundary to the Google search. The study entails the improvement of Hands- Free voice recognition Google Search Engine to operate Google and browse the result of search without using a keyboard or mouse. Speech recognition uses are becoming more and more beneficial nowadays. Digital processing of speech signal and voice recognition process is very important for fast and precise automatic voice recognition technology. Here we present a new service which is not currently accessible in Google search engine (GSE). It suggests the enactment of speech recognition input in GSE. The paper stimulates an older method from n-gram language modelling to scale training data. The algorithm is implemented efficiently using a MapReduce/SS (Spectral Subtraction) Table framework based on HMM and Gaussian Mixture models.