Vowel recognition system of Lipsynchrobot in lips gesture using neural network

摘要

In this research, we propose a system of vowel recognition on the shape of the lips using a neural network backpropagation. For recognizing the shape of the vowels in the lips by image capturing through the webcam, and then processed through image processing which includes edge detection, filtering, mouth feature extraction, integral projection and vowels recognition on the lips using a back propagation neural network. The vowel recognition results obtained at the lips of the testing process neural network based on the training data and test data. The proposed method works well. The final results obtained from this research is a system that able to detect the lips with good features, with the success of recognizing vowels using 250 training data is 70%. The output data from the vowel recognition on the lips is used as an input to Lipsynchrobot that followed the gesture of human lips.