Why
-
If you can’t speak in a classroom, venue, etc., you want to interact without speaking out
-
Tell the story without pronunciation
Background
-
Researchers at Oxford University and Google DeepMind have developed artificial intelligence (AI) that reads the movement of a person’s lips through training with thousands of hours of content broadcast by the BBC.
-
Little research on facial muscle activity
-
Difficult to discriminate micro electromyographic activity
-
Obtained action potential is a continuous waveform that fluctuates in time and is difficult to analyze usually.
How
Different mouth shape when speaking different vowels
By collecting masticatory EMG data when speaking different vowels(chin–plus,cheek–minus,forehead–reference)
Result
-
I analyzed the EMG data by analyzing envelope of different signals, different signals can be easily distinguished
-
By using a simple neural network from MATLAB, different EMG signals of different vowels can be correctly distinguished.
