Abstract

The science and study of speech recognition has been around for over half a century, but it is only in the last decade or so that we have seen a huge proliferation of Automated Speech Recognition technology. The Hidden Markov Model is just as old, and one of its first applications was, fittingly so, for Speech Recognition. Despite the appearance of other Speech Recognition techniques, the Hidden Markov Model has withstood the test of time and remains as the most widely used approach for modern general-purpose speech recognition systems.

In this talk we shall explore from first principles what speech is, before delving into Hidden Markov Models and the mathematical tools we will need for Speech Recognition. We will look at how we can practically apply these tools in computer algorithms and lastly take the step of turning acoustic signals into the meaningful words we are all familiar with.