A perceptual dissimilarities based nonlinear sound quality model for range hood noise

Area of Science:

• Acoustics
• Psychoacoustics
• Artificial Intelligence

Background:

• Growing importance of sound quality in household appliances.
• Need for appropriate acoustic metrics alongside modeling algorithms.
• Limited interpretability and nonlinearity modeling in existing sound quality models.

Purpose of the Study:

• To establish an artificial neural network-based sound quality model for range hood noise.
• To integrate multidimensional scaling (MDS) for metric selection.
• To enhance the interpretability and accuracy of sound quality models.

Main Methods:

• Recording and subjective evaluation of range hood noise annoyance.
• Multidimensional scaling (MDS) analysis using CLASCAL algorithm to analyze timbre space.
• Stepwise regression for metric selection and back propagation neural network (BPNN) for model development.
• Application of passive and active noise control (ANC) measures.

Main Results:

• Successful establishment of a sound quality model combining BPNN and CLASCAL.
• Demonstrated ability to interpret the sound quality model and model nonlinearity.
• Validation of passive and active noise control (ANC) effectiveness in improving range hood sound quality.

Conclusions:

• The combination of BPNN and CLASCAL offers high accuracy and interpretability for sound quality modeling.
• Active noise control (ANC) is particularly effective in enhancing the sound quality of range hoods.
• The developed model provides a robust framework for assessing and improving appliance acoustics.