摘要 |
<p>In a speech processor such as a speech recogniser, the problem of detecting the beginning and end of speech or a word accurately, to enable the creation of a speech or a word template which consistently matches stored speech or word templates is solved by characterising background noise and forming a background noise template, setting a speech threshold above which speech is detected and stored, and subtracting the background noise template from the stored speech to form a speech template. <IMAGE></p> |