Building speaker recognition module for VNPT ecabinet using deep learning technique | |
| Author | Vu Minh Duc |
| Call Number | AIT Project no.PMDS-23-02 |
| Subject(s) | Automatic speech recognition Deep learning (Machine learning) |
| Note | A project study submitted in partial fulfillment of the requirements for the degree of Professional Master in Data Science and Artificial Intelligence Applications |
| Publisher | Asian Institute of Technology |
| Abstract | Speaker Recognition is an added value in the paperless meeting system as VNPT eCabinet, where it can release the effort of help desk staff from listening to recognize and tag speaker names in recorded content of meetings. Moreover, in the AI era, smart features like this are the keys to impress customers and win the bid in Vietnam. This project aims to apply deep learning techniques for building an added value module for VNPT eCabinet - a speaker recognition module, in which we compared the accuracy of two DNN models: original RNN and LSTM; using MFCC to extract features from speeches in recognizing speaker tasks. An experiment was conducted with VIVOS Corpus – free Vietnamese speeches dataset of 65 speakers with 12420 utterances. Our key findings are that LSTM Bidirectional with EER = 0.136 is better to resolve this problem than original RNN (EER = 0.210). Our work could help VNPT to leverage machine learning models to reduce help desk efforts and increase the competition of VNPT eCabinet with other Vietnam paperless meeting systems. |
| Year | 2023 |
| Type | Project |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Professional Master in Data Science and Artificial Intelligence Applications (PMDS) |
| Chairperson(s) | Chaklam Silpasuwanchai |
| Examination Committee(s) | Vatcharaporn Esichaikul;Chantri Polprasert |
| Degree | Professional Master in Data Science and Artificial Intelligence Applications - Asian Institute of Technology, 2023 |