|
Digital Thesis Room >
Faculty of Graduate Studies >
Theses & Dissertations >
Please use this identifier to cite or link to this item:
http://hdl.handle.net/10791/369
|
| Title: | AUTOMATED SPOKEN LANGUAGE DETECTION |
| Authors: | Pennell, Ripley |
| Supervisor(s): | Dr. Maiga Chang, Athabasca University |
| Examining Committee: | Dr. Ali Dewan, Athabasca University
Dr. Kuo-Chen Li, Chung-Yuan Christian University |
| Degree: | Master of Science, Information Systems (MScIS) |
| Department: | Faculty of Science and Technology |
| Keywords: | Natural Language Processing
i-vector
Language Identification (LID)
Automatic Speech Recognition (ASR)
Kaldi
Mozilla Common Voice
TensorFlow
Chaqoupy |
| Issue Date: | 19-Jan-2022 |
| Abstract: | This research allows two individuals to speak their language with an application detecting what languages are being spoken, allowing automatic translation. Existing relevant Systematic Literature Reviews (SLRs) articulated the need for this research. An SLR with quantitative and qualitative analysis identified the best algorithm to use, the i-vector algorithm. To integrate it onto a mobile platform it had to be completely recreated, referencing Kaldi. A voice database was created using Mozilla Common Voice and four (4) models were trained using TensorFlow, each showing unique improvements. The final model is deployed in an Android application using Chaqoupy for environment translation. Evaluation produced an accuracy of 81% and a 95.7 on the System Usability Scale. Evaluation data was transformed for normality and analyzed using a one-way analysis of variance and a two independent samples t-test. This research can be applied to all languages and has no dependency on accents. |
| Graduation Date: | Jun-2022 |
| URI: | http://hdl.handle.net/10791/369 |
| Appears in Collections: | Theses & Dissertations
|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.
|
