| Autor: |
Taro Kasahara, Hikaru Watabe, Taichi Ikeda, Hiroshi Yoshikoshi |
| Zdroj: |
Transactions of the Society of Automotive Engineers of Japan; Jun2022, Vol. 53 Issue 4, p717-722, 6p |
| Abstrakt: |
The deep learning model (DNN, Deep Neural Net) we proposed separates knocking sound from engine radiation noise measured by a microphone. In this paper, we propose a method for training a DNN which can be used in a wide range of rotational speed conditions. The proposed method applies data augmentation to measurement data under several conditions. In addition, the proposed method trains a DNN (UNet) for sound source separation using the results of the previous method. The proposed method can reduce the time required for training data collection. In terms of practical application, reducing the burden of data collection is an important improvement. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
