Speaker recognition can be done effectively using MFCC features and fitting a Gaussian Mixture Model. In our project we had the data from 22 different speakers. Each speaker were having 8 voice samples for the training. A GMM model was fit for each speaker. When a new test voice comes the likelyhood of the voice is computed for all the speaker models and the one with the highest likelyhood will identified as the speaker.
A detail description and codes are present in the following link
Speaker recognition matlab code
A detail description and codes are present in the following link
Speaker recognition matlab code
No comments:
Post a Comment