Attention Makers

CATEGORY: HEALTH & SAFETY

Speaker Identification using Deep learning

MAKERS: Krishna , Noor Fathima Khanum COUNTRY: India

This project is to teach machines to recognize the speech of people using Deep learning method

The Purpose

This project can be used biometric system to indentify the person through his/her voice. Here we use Deep Neural networks to authenticate the person identity because they give very good performance.

The Technology

We use Deep Neural network based approach to identify the speaker.The network is presented with the MFCC features of a specific speaker along with the label which says who is speaking. After training, the network would have learnt the speakers charecteristics. Now when one of the speaker in the list speaks again the network will know who is speaking and make the decision.

Additional Details

Here we train DNN for 4 different speakers with less number of layers to avoid overfitting. we use 13 dimensional MFCC along wth delta coefficiants and also with 10 frames of left and right context

Home Previous Next

Vote Share Comment