Intelligent classification of automotive horn sound quality

The primary function of an automotive horn is to alert pedestrians and other nearby vehicles to their safe passage on the road. Most of the human population is subjected to a certain amount of horn sound dosage daily. The study of automotive horns sound quality is equally important as their sound generation mechanism in high-end cars. The sound quality of automotive horns can be studied through subjective and objective test methods. In the present study, a subjective jury test and objective analysis using psychoacoustic parameters are conducted to classify car horn sound samples according to pleasantness. A total of twenty-two car horns, which consist of a disc and shell, are chosen for binaural sound recording. The recorded sound samples are used for subjective and objective analysis. Thirty members participated in the jury test, and a semantic differential method was selected to collect the user response. The Tukey range test is used to classify the subjective test data into groups. Six parameters, namely SPL (dBA), loudness, sharpness, roughness, fluctuation strength, and tonality, are obtained from the recorded sound samples of automotive horns using in-house-developed code for objective analysis. The calculated objective parameters are further used to group them by Mahalanobi's distance method and AI/ML based intelligent classification techniques. Critical psychoacoustic parameters with the highest weightage on the classification of horn sound samples were identified and AI/ML models reconstructed. Reconstructed models produce similar classification accuracy as compared to the original AI models. Horn sound classifications based on subjective tests, objective analysis, and intelligent classification techniques are found to be in good agreement. Developed models can be used for the selection of a pleasant class of automotive horn for final deployment.