Prediction of Detectability of Synthesized Vehicle Sounds Using Logistic Regression

Design of the alert sound for the electric vehicles includes not only esthetical concerns but also safety aspects. The timely detection of the vehicles by pedestrians is a prerequisite for road safety. To determine the detectability characteristics of the vehicle sounds, usually psychoacoustical investigation with test participants will be conducted. To minimize the effort, which is necessary for listening experiments, it is helpful to develop a quantitative index for detectability of vehicle exterior sounds. In this study, a model, which is based on logistic regression, was developed. The criteria for the model is the detection time. It was investigated in function of various variables. As independent variables, A weighted sound pressure level, loudness, sharpness, tonality, roughness, fluctuation strength, frequency (third octave bands) and their number are defined. The logistic regression allows us to make a statement about the detectability characteristics of the synthesized vehicle sounds.