| With more devices installed at home and the stronger desire for a more comfortable and easier controllable home environment, the ability to use voice as the unified control method to resolve the confusion of multiple user interfaces looks appealing. This thesis first provides the reader with necessary background of home automation and speech recognition technologies. Then a detailed procedure on how to use Hidden Markov Toolkit (HTK) to build a Hidden Markov Model (HMM) based speech recognizer is introduced. The robustness of the speech recognizer is tested with various types of noises that typically exist in a home environment. It is shown that using voice as the control method can produce high recognition accuracy as long as limited vocabulary and restricted grammar are used. Finally, the speech recognizer is integrated with an X.10 based light controller to demonstrate the feasibility of voice activation for home devices. |
