SPEAKER RECOGNITION SUB-SYSTEM USING MFCC AND DATABASE SQL FOR TITAN (TEACHING ASSISTANT)

• Main Author: ARY - NIM : 13214083 , JOHANES
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/28289
• Institution: Institut Teknologi Bandung

<p align="justify">This final project describes about the digital signal processing for speech to know the identity of speaker. This is a part of Teaching Asisstant (TITAN) sub sytem that will be implemented in smart class room. This feature will be useful to give permission for teacher to give a specific command to TITAN. Feature extraction use MFCC method and then we do classification of feature using VQ. The training data of system was taken in semi anechoic chamber (sound proof) so the the data was clear and free from noise from outside or sound reflection caused by wall. Total amount of data training is 1080 samples with 93 different speakers. Testing was done to know the accuracy of sytem, threshold limit for noise, and time execution. The result is system has 71% accuracy with 60 dB threshold limit for environmental noise, and less than 2 seconds for time execution. After we get all of information from speech data, we need to save it in database as a history or future development. Then we show it in the web. <p align="justify">