Abstract HTK is a speech signal processing toolkit developed by CUED based on C language. It is widely used for speech signal recognition，

speech synthesis，character recognition and so on. This paper describes the process and principles of speech recognition and related

concepts. Then，the basic principles and software architecture of HTK are introduced. The entire process of speech recognition task is completed

with HTK. Finally，a chinese continuous speech input system was built and discuss how to select a series of acoustic unit model parameter.