Human Hearing Based Model for Workshop Noise Prediction

ABSTRACT With the growing concern over the impact of workshop noise on workers' health, a dependable method to predict the workshop noise has become imperative. This study introduces a prediction model for workshop noise, which combines human hearing and bidirectional long short‐term memory (BiLSTM) neural networks. First, the collected noise signal is divided into 24 critical bands according to frequency. The time‐varying specific loudness is then calculated. The division of critical bands is consistent with the channel divided by human hearing. Therefore, the critical bands reflects the basic properties of human hearing, and this prediction process also conforms to human auditory perception. Finally, BiLSTM neural network is used to predict each critical band individually. The mean absolute error (MAE), the mean absolute percentage error (MAPE), the root mean square error (RMSE), the peak signal‐to‐noise ratio (PSNR), and the structural similarity index measure (SSIM) of the model proposed in this paper are 0.7328, 8.61%, 1.1839, 19.688, and 93.77%, respectively. This paper also compares the proposed model with several commonly used models, including Empirical Mode Decomposition (EMD), Variational Mode Decomposition (VMD), and others. The experiment proved that the proposed approach exhibits excellent accuracy and high efficiency.