This content is not included in your SAE MOBILUS subscription, or you are not logged in.
Modeling Subjective Evaluation of Instantaneous Sound Qualities of Motorcycle Exhaust Sound Applying a Highly Efficient Experimental Design
ISSN: 2641-9637, e-ISSN: 2641-9645
Published January 24, 2020 by Society of Automotive Engineers of Japan in Japan
Citation: Tanaka, K., Nishina, S., Sugita, H., Kato, T. et al., "Modeling Subjective Evaluation of Instantaneous Sound Qualities of Motorcycle Exhaust Sound Applying a Highly Efficient Experimental Design," SAE Int. J. Adv. & Curr. Prac. in Mobility 2(2):1058-1066, 2020.
Exhaust sound quality is one of the most important factors for attractive motorcycles. It is desirable to quantitatively and objectively examine the exhaust sound quality during the development process.
In our previous study, we developed evaluation models that could be applied to various exhaust sounds. However, due to the limitation of the model, it could not be applied to varying conditions, such as from acceleration to steady driving and then to deceleration. In this study, we developed more practical evaluation models that can be applied to various driving conditions.
To cover a wide variety of sounds under different driving conditions, we prepared more than 300 sound stimuli and collected subjective evaluation data on those sounds. As the evaluation method, we used a pairwise comparison method to make decision making easy for the participants. However, if the pairwise comparison was simply performed for such sample size, the experiment cannot be done in a realistic experimental scale because the number of evaluation pairs gets very large. To solve this issue, we allocated evaluating pairs among multiple participants and used the Bradley-Terry model for estimation of the subjective quality rate. With these methods, the number of evaluations per participant was reduced to 1/300 of that required with the naive method.
We constructed the models based on a multiple regression analysis on the subjective evaluation data with waveform feature values as independent variables. We confirmed that these models accurately predicted the motorcycle developers' subjective impression of the exhaust sound of actual vehicles under a wide range of driving conditions.