Analysis of correlation between window duration for kurtosis computation and accuracy of noise-induced hearing loss prediction

A Noise-Induced Hearing Loss Prediction Model Based on Asymmetric Convolution for Workers Exposed to Complex Industrial Noise

OBJECTIVES

Current approaches for evaluating noise-induced hearing loss (NIHL), such as the International Standards Organization 1999 (ISO) 1999 prediction model, rely mainly on noise energy and exposure time, thus ignoring the intricate time-frequency characteristics of noise, which also play an important role in NIHL evaluation. In this study, an innovative NIHL prediction model based on temporal and spectral feature extraction using an asymmetric convolution algorithm is proposed.

DESIGN

Personal data and individual occupational noise records from 2214 workers across 23 factories in Zhejiang Province, China, were used in this study. In addition to traditional metrics like noise energy and exposure duration, the importance of time-frequency features in NIHL assessment was also emphasized. To capture these features, operations such as random sampling, windowing, short-time Fourier transform, and splicing were performed to create time-frequency spectrograms from noise recordings. Two asymmetric convolution kernels then were used to extract these critical features. These features, combined with personal information (e.g., age, length of service) in various configurations, were used as model inputs. The optimal network structure was selected based on the area under the curve (AUC) from 10-fold cross-validation, alongside the Wilcoxon signed ranks test. The proposed model was compared with the support vector machine (SVM) and ISO 1999 models, and the superiority of the new approach was verified by ablation experiments.

RESULTS

The proposed model had an AUC of 0.7768 ± 0.0223 (mean ± SD), outperforming both the SVM model (AUC: 0.7504 ± 0.0273) and the ISO 1999 model (AUC: 0.5094 ± 0.0071). Wilcoxon signed ranks tests confirmed the significant improvement of the proposed model ( p = 0.0025 compared with ISO 1999, and p = 0.00142 compared with SVM).

CONCLUSIONS

This study introduced a new NIHL prediction method that provides deeper insights into industrial noise exposure data. The results demonstrated the superior performance of the new model over ISO 1999 and SVM models. By combining time-frequency features and personal information, the proposed approach bridged the gap between conventional noise assessment and machine learning-based methods, effectively improving the ability to protect workers' hearing.

Investigation of critical factors influencing the underestimation of hearing loss predicted by the ISO 1999 predicting model

OBJECTIVE

To analyze factors influencing the underestimation of noise-induced permanent threshold shift (NIPTS) among manufacturing workers, providing baseline data for revising noise exposure standard.

DESIGN

A cross-sectional study was designed with 2702 noise-exposed workers from 35 enterprises from 10 industries. Personal noise exposure level(LAeq,8h) and noise kurtosis level were determined by a noise dosimeter. Questionnaires and hearing loss tests were performed for each subject. The predicted NIPTS was calculated using the ISO 1999:2013 model for each participant, and the actual measured NIPTS was corrected for age and sex. The factors influencing the underestimation of NIPTS were investigated.

RESULTS

The predicted NIPTS at each test frequency (0.5, 1, 2, 3, 4, or 6kHz) and mean NIPTS at 2, 3, 4, and 6kHz (NIPTS2346) using the ISO 1999:2013 model were significantly lower than their corresponding measured NIPTS, respectively (P < 0.001). The ISO model significantly underestimated the NIPTS2346 by 12.36 dB HL. The multiple linear regression analysis showed that noise exposure level, exposure duration, age, and kurtosis could affect the degree of underestimation of NIPTS2346. The generalized additive model (GAM) with (penalized) spline components showed nonlinear relationships between critical factors (age, exposure duration, noise level, and kurtosis) and the underestimated NIPTS2346.The underestimated NIPTS2346 decreased with an increase in exposure duration (especially over ten years). There was no apparent trend in the underestimated NIPTS2346 with age. The underestimated NIPTS2346 decreased with the increased noise level [especially > 90 dB(A)]. The underestimated NIPTS2346 increased with an increase in noise kurtosis after adjusting for the noise exposure level and exposure duration and ultimately exhibiting a linear regression relationship.

CONCLUSIONS

The ISO 1999 predicting model significantly underestimated the noise-induced hearing loss among manufacturing workers. The degree of underestimation became more significant at the noise exposure condition of fewer than ten years, less than 90 dB(A), and higher kurtosis levels. It is necessary to apply kurtosis to adjust the underestimation of hearing loss and consider the applying condition of noise energy metrics when using the ISO predicting model.

Developing a guideline for measuring workplace non-Gaussian noise exposure based on kurtosis adjustment of noise level in China

Objective

There is no unified standard for measuring workplace non-Gaussian noise (known as complex noise) exposure. This study aimed to develop a draft guideline for measuring workplace non-Gaussian complex noise exposure based on noise temporal structure adjustment.

Methods

Noise exposure level, e.g., the A-weighted sound pressure level normalized to a nominal 8-h working day (L_EX,8h), was adjusted using the temporal structure (expressed by kurtosis) of noise. Noise waveform analysis or the instrument's direct reading was used.

Results

The framework of the draft guideline included measurement metrics, the protocol using kurtosis to adjust L_EX,8h, technical requirements for measuring instruments, measurement steps, data analysis, and measurement recording.

Conclusion

The draft guideline could provide a basis for accurately measuring workers' exposure to non-Gaussian noise.

Evaluation of kurtosis-corrected sound exposure level as a metric for predicting onset of hearing threshold shifts in harbor porpoises (Phocoena phocoena)

Application of a kurtosis correction to frequency-weighted sound exposure level (SEL) improved predictions of risk of hearing damage in humans and terrestrial mammals for sound exposures with different degrees of impulsiveness. To assess whether kurtosis corrections may lead to improved predictions for marine mammals, corrections were applied to temporary threshold shift (TTS) growth measurements for harbor porpoises (Phocoena phocoena) exposed to different sounds. Kurtosis-corrected frequency-weighted SEL predicted accurately the growth of low levels of TTS (TTS_1-4 < 10 dB) for intermittent sounds with short (1-13 s) silence intervals but was not consistent with frequency-weighted SEL data for continuous sound exposures.

Estimation of Occupational Noise-Induced Hearing Loss Using Kurtosis-Adjusted Noise Exposure Levels

OBJECTIVES

Studies have shown that in addition to energy, kurtosis plays an important role in the assessment of hearing loss caused by complex noise. The objective of this study was to investigate how to use noise recordings and audiometry collected from workers in industrial environments to find an optimal kurtosis-adjusted algorithm to better evaluate hearing loss caused by both continuous noise and complex noise.

DESIGN

In this study, the combined effects of energy and kurtosis on noise-induced hearing loss (NIHL) were investigated using data collected from 2601 Chinese workers exposed to various industrial noises. The cohort was divided into three subgroups based on three kurtosis (β) levels (K 1 : 3 ≤ β ≤ 10, K 2 : 10 <β ≤ 50, and K 3 : β > 50). Noise-induced permanent threshold shift at test frequencies 3, 4, and 6 kHz (NIPTS 346 ) was used as the indicator of NIHL. Predicted NIPTS 346 was calculated using the ISO 1999 model for each participant, and the actual NIPTS was obtained by correcting for age and sex using non-noise-exposed Chinese workers (n = 1297). A kurtosis-adjusted A-weighted sound pressure level normalized to a nominal 8-hour working day (L Aeq,8h ) was developed based on the kurtosis categorized group data sets using multiple linear regression. Using the NIPTS 346 and the L Aeq.8h metric, a dose-response relationship for three kurtosis groups was constructed, and the combined effect of noise level and kurtosis on NIHL was investigated.

RESULTS

An optimal kurtosis-adjusted L Aeq,8h formula with a kurtosis adjustment coefficient of 6.5 was established by using the worker data. The kurtosis-adjusted L Aeq,8h better estimated hearing loss caused by various complex noises. The analysis of the dose-response relationships among the three kurtosis groups showed that the NIPTS of K 2 and K 3 groups was significantly higher than that of K 1 group in the range of 70 dBA ≤ L Aeq,8h < 85 dBA. For 85 dBA ≤ L Aeq,8h ≤ 95 dBA, the NIPTS 346 of the three groups showed an obvious K 3 > K 2 > K 1 . For L Aeq,8h >95 dBA, the NIPTS 346 of the K 2 group tended to be consistent with that of the K 1 group, while the NIPTS 346 of the K 3 group was significantly larger than that of the K 1 and K 2 groups. When L Aeq,8h is below 70 dBA, neither continuous noise nor complex noise produced significant NIPTS 346 .

CONCLUSIONS

Because non-Gaussian complex noise is ubiquitous in many industries, the temporal characteristics of noise (i.e., kurtosis) must be taken into account in evaluating occupational NIHL. A kurtosis-adjusted L Aeq,8h with an adjustment coefficient of 6.5 allows a more accurate prediction of high-frequency NIHL. Relying on a single value (i.e., 85 dBA) as a recommended exposure limit does not appear to be sufficient to protect the hearing of workers exposed to complex noise.

Kurtosis loss as a metric for hearing protection evaluation in impulsive noise environments

Hearing loss standards depend on noise power and duration but are incomplete when the noise is primarily impulsive in nature rather than maintaining a continuous power level. Calculating the kurtosis of a noise exposure captures information about its impulsivity, and high kurtosis values cause additional hearing damage. In this paper, a method for measuring the reduction of noise kurtosis through hearing protection is outlined, and measurements demonstrate that spectral insertion loss is independent of the noise kurtosis and that kurtosis loss is not related to either the mean or standard deviation of spectral attenuation.