PENGENALAN SUARA KE TEKS MENGGUNAKAN GAUSSIAN MIXTURE MODEL
Total View This Week0

Institusion

Universitas Sriwijaya

Author

FATARAH, MUHAMMAD DAFFA RIZKY
Arsalan, Osvari
Kurniati, Rizki

Subject

Q325.5 Machine learning 

Datestamp

2023-04-12 02:37:58 

Abstract :
English content is very easy to find in the surrounding environment such as in social media, but Indonesians find it difficult to understand. Since English word often sounds similar to other words and also the pronunciation differences of each person, a speech-to-text recognition system can help recognize English words into text. Gaussian Mixture Model is chosen to recognize speech to text because it is better than other speech recognition machine learning methods. The stages in GMM include Lower Bound and Expectation Maximization, where the result of GMM is a Gaussian distribution with mean (?) and variance (?2) parameters. Therefore, a voice-to-text recognition application was developed in this research using the Gaussian Mixture Model method. Three GMM models were created using three different datasets and the same parameter configuration. The best model gives perfect results with 100% accuracy using datasets that have stable voice signals, clear pronunciation, and little test data. 

File :
RAMA_55201_09021181924014.pdf
RAMA_55201_09021181924014_TURNITIN.pdf
RAMA_55201_09021181924014_0028068806_0012079104_01_front_ref.pdf
RAMA_55201_09021181924014_0028068806_0012079104_02.pdf
RAMA_55201_09021181924014_0028068806_0012079104_03.pdf
RAMA_55201_09021181924014_0028068806_0012079104_04.pdf
RAMA_55201_09021181924014_0028068806_0012079104_05.pdf
RAMA_55201_09021181924014_0028068806_0012079104_06.pdf
RAMA_55201_09021181924014_0028068806_0012079104_07_ref.pdf
RAMA_55201_09021181924014_0028068806_0012079104_08_lamp.pdf