Speech enhancement using artificial neural network
This dissertation will investigate various methods of noise reduction in speech signals using back propagation neural networks. Neural network approach on time domain and transform domain methods are focused. A comprehensive evaluation and comparison on the performance of both time and transform dom...
Saved in:
| 主要作者: | |
|---|---|
| 其他作者: | |
| 格式: | Theses and Dissertations |
| 語言: | English |
| 出版: |
2014
|
| 主題: | |
| 在線閱讀: | http://hdl.handle.net/10356/55246 |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 機構: | Nanyang Technological University |
| 語言: | English |
| 總結: | This dissertation will investigate various methods of noise reduction in speech signals using back propagation neural networks. Neural network approach on time domain and transform domain methods are focused. A comprehensive evaluation and comparison on the performance of both time and transform domain approach is carried out. In precise network architecture, training issues and efficiency in noise reduction are investigated. In addition, speech enhancement techniques like magnitude spectral subtraction are explored and their results are discussed.
Finally, the results of both time and transform domain filtering are compared and tabulated. Artificially corrupted clean speech at different SNR levels is used as test set. Objective tests are done based on signal-to-noise ratios and segmental signal-to-noise ratios. The strengths and weakness of both time domain and transform domain mapping techniques are analyzed and compared. |
|---|