Efficiency analysis of acoustic speech recognition unit with different noises

Abstract

One of the most important criteria in human-machine communication is reliability. This expectation includes that the detected information be the same as the information content of the signal emitted by the sender. We know that one of the most basic means of communication between people and at the same time the simplest and most natural way of doing so is speaking. With this statement in mind, it is man's old desire to be able to communicate effectively and reliably with the machines and devices he constructs in human language, with speech. During the course of my work, I will perform tests on a speech recognizer based on hidden Markov models. I will strive to select the most optimal number of states for the hidden Markov models, and then I will investigate the effects of different noise-laden teaching and testing materials on efficiency.