Small-scale model for classification of single acoustic events
This project was primarily about exploring the use of real-world and noisy datasets for sound event classification. To expand the capabilities of machine learning models in their real-world applications, learning models must be trained on real-world data. However, using real-world data comes with it...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2021
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/150261 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | This project was primarily about exploring the use of real-world and noisy datasets for sound event classification. To expand the capabilities of machine learning models in their real-world applications, learning models must be trained on real-world data. However, using real-world data comes with its own set of challenges and the most apparent one is the effect of noise on the performance of sound event recognition systems built on said models. Therefore, this project aimed to understand those challenges and limitations through the process of designing a learning model that performs the task of classifying sound events occurring in real-world environment well.
Although the model created for the purpose of this project had not been fully optimized and had subpar performance relative to the model from which it was adapted, it demonstrated (to a limited extent) the potential and possibility of using simple models to tackle some of the challenges of sound event recognition in complex environments.
Finally, towards the end of the report, the difficulties encountered in this project and how this project can be expanded on were discussed. |
|---|