This research developed a unique process for the automatic classi- fication of speech signals. Instead of classifying speech signals into common categories such as who the speaker is or what language is being spoken, this research examines ways to classify the speech into three broad classes of speech production - normally phonated speech, whispered speech, and softly spoken speech. Since most automated speech processing techniques give erroneous results when presented with a mixture of these classes of speech, this technique can provide class information to improve automated speech processing systems performance.