A sound quality model for objective synthesis evaluation of vehicle interior noise based on artificial neural network

• Some objective sound quality evaluation (SQE) methods for vehicle interior noises are discussed.
• Sound annoyance of interior noises is estimated by jury tests following the anchored semantic differential procedure.
• A model for annoyance evaluation of vehicle noises, so-called ANN-SAE, is developed and verified.
• The ANN-SAE can synthetically evaluate sound annoyance, which may be used as a substitution in SQE engineering.
• In applications, the ANN-SAE approach is helpful for vehicle acoustical designs and improvements in SQE engineering.

Based on the artificial neural network (ANN) technique, an objective sound quality evaluation (SQE) model for synthesis annoyance of vehicle interior noises is presented in this paper. According to the standard named GB/T18697, firstly, the interior noises under different working conditions of a sample vehicle are measured and saved in a noise database. Some mathematical models for loudness, sharpness and roughness of the measured vehicle noises are established and performed by Matlab programming. Sound qualities of the vehicle interior noises are also estimated by jury tests following the anchored semantic differential (ASD) procedure. Using the objective and subjective evaluation results, furthermore, an ANN-based model for synthetical annoyance evaluation of vehicle noises, so-called ANN-SAE, is developed. Finally, the ANN-SAE model is proved by some verification tests with the leave-one-out algorithm. The results suggest that the proposed ANN-SAE model is accurate and effective and can be directly used to estimate sound quality of the vehicle interior noises, which is very helpful for vehicle acoustical designs and improvements. The ANN-SAE approach may be extended to deal with other sound-related fields for product quality evaluations in SQE engineering.