Development of a Machine Learning Model to Predict Engine Noise Perception Considering Temporal Driving Conditions
走行状態の時間履歴有無を考慮したエンジン音評価推定モデルの構築
- Delivery
- Available on the other site
- Click here to order.
- Publication code
- 20264196
- Paper/Info type
- JSAE Transaction
Vol.57 No.2
- Pages
- 374-380(Total 7 p)
- Date of publication
- Mar 2026
- Publisher
- JSAE
- Language
- Japanese
Detailed Information
| Category(J) | 論文 Translation |
|---|---|
| Category(E) | Paper |
| Author(J) | 1) 菅沼 真一, 2) 長江 新平, 3) 戸井 武司 |
| Author(E) | 1) Suganuma Shinichi, 2) Nagae Shimpei, 3) Toi Takeshi |
| Affiliation(J) | 1) 中央大学大学院，日産自動車株式会社, 2) 日産自動車株式会社, 3) 中央大学 |
| Abstract(J) | エンジン音主観評価を推定する機械学習モデルにおいて，主観評価に寄与の高いエンジン回転数，ドライブシャフトトルク，車速を説明変数とした場合の推定精度65 %に対し，評価直前5.5秒間のエンジン回転数履歴を追加することで精度は75 %へ改善し，主観評価に対する時間要因の影響を定量的に示した． Translation |
| Abstract(E) | In hybrid vehicles, the acoustic contrast between engine-off and engine-on states increases annoyance from engine noise. Machine learning models were developed to quantify subjective engine noise perception based on synchronized in-cabin sound and vehicle signals. A random forest classifier trained on engine speed, driveshaft torque, and vehicle speed achieved 65 % accuracy, further improving to 75 % by including a 5.5-second engine speed history. The model quantitatively demonstrated that annoyance increases when engine speed is rising and decreases when it is falling, even at the same instantaneous speed, showing that temporal changes are important for noise perception and control. |
