|
1 | 1 | # Analysis of Audio Signals Using Linear Predictive Coding
|
2 | 2 | This study deals with audio signal feature extraction in order to be used for speaker authentication using a neuronal network.
|
3 | 3 | Specifically, the effectiveness of linear predictive coding (LPC) coefficients is examined.
|
4 |
| -The goal of this study is to explain how linear predictive coefficients can be extracted and to evaluate whether they can be used to differentiate between multiple speakers. |
| 4 | +The goal of this study is to explain how LPC coefficients can be extracted and to evaluate whether they can be used to differentiate between multiple speakers. |
| 5 | +Therefore, the developed audio preprocessing (noise and silence removal, framing and windowing) and LPC extraction method is applied to samples of 10 speakers from the [data set](https://www.kaggle.com/datasets/vjcalling/speaker-recognition-audio-dataset?resource=download). |
| 6 | +A simple neural network is then trained and tested with the extracted features. |
| 7 | + |
| 8 | +## Results |
| 9 | +The evaluation of the data set using the neural network resulted in a prediction accuracy of **70.54 percent**, showing a loss of 5.47. |
| 10 | +Thus the effectiveness of LPC for speaker authentication is proven. |
| 11 | + |
| 12 | +## Subsequent studies |
| 13 | +### User authentication using voice recognition |
| 14 | +[](https://github.com/DHBW-FN-TIT20/sa-hs-lb-jb)</br> |
| 15 | +The results of this study form the basis for the subsequent student research project. |
| 16 | +Within the project, LPC is combined with other speaker related audio features like mel frequency cepstral coefficients to create a neuronal network structure that is capable of authenticating speakers. |
| 17 | +The main goal of the student research project is to improve the systems accuracy by variating the calculated coefficients as well as the structure of the neural network. |
5 | 18 |
|
6 | 19 | ## Author
|
7 |
| -* [Henry Schuler](https://henryschuler.de) / [github](https://github.com/schuler-henry) / [E-Mail](mailto:contact@henryschuler.de?subject=[GitHub]%20dhbw-latex-template) |
| 20 | +### Henry Schuler |
| 21 | +[](https://github.com/schuler-henry) |
| 22 | +[](mailto:contact@henryschuler.de?subject=[GitHub]%20analysis-of-audio-signals-using-linear-predictive-coding) |
8 | 23 |
|
9 | 24 | ## [LICENSE](LICENSE)
|
10 | 25 | Copyright (c) 2022 Henry Schuler
|
|
0 commit comments