TY - JOUR
T1 - An adaptive audio injection method based on dual-time-scale cooperative control
AU - Luo, Jianfeng
AU - Chen, Kean
AU - Li, Hao
AU - Deng, Yunyun
AU - Liu, Fancheng
AU - Zhao, Huanqi
N1 - Publisher Copyright:
© 2026 Elsevier Ltd.
PY - 2027/2/5
Y1 - 2027/2/5
N2 - The growing concern over product noise has shifted the focus from acoustic energy to perceived annoyance. Traditional noise control techniques are limited in this case, but the audio injection method (AIM) reduces perceived annoyance by injecting a controllable sound into the original noise (target sound). However, AIM performance heavily depends on the selection and parameterization of the controllable sound, and existing approaches lack adaptability in non-stationary conditions. This study proposes an adaptive audio injection method (AAIM) based on a dual-time-scale control framework, featuring a cooperative strategy that integrates long-term frame band energy distribution selection with short-term frame subband gain optimization. At the long-term scale, a perceived annoyance prediction model quantitatively evaluates each critical band’s contribution to the target sound’s total annoyance, and an automatic selection mechanism chooses a controllable sound from candidate sounds that balances spectral similarity with a minimal sound pressure level (SPL) redundancy index. At the short-term scale, the simultaneous perturbation stochastic approximation gradient-descent (SPSA-GD) algorithm iteratively optimizes the gains of each critical band using annoyance as the cost function, enabling refined spectral reconstruction. Furthermore, to enhance system responsiveness, performance is analyzed and a variable step-size strategy is introduced, forming the VSS AAIM. The proposed methods are validated through single-channel simulations and experiments using range hood noise. Under hybrid operating condition, switching from strong turning to quick-fry mode, the VSS AAIM reduced actual annoyance by 1.01 and 1.16 compared to the fixed AIM, without a significant increase in SPL, achieving a perceptible improvement and demonstrating the method’s superiority.
AB - The growing concern over product noise has shifted the focus from acoustic energy to perceived annoyance. Traditional noise control techniques are limited in this case, but the audio injection method (AIM) reduces perceived annoyance by injecting a controllable sound into the original noise (target sound). However, AIM performance heavily depends on the selection and parameterization of the controllable sound, and existing approaches lack adaptability in non-stationary conditions. This study proposes an adaptive audio injection method (AAIM) based on a dual-time-scale control framework, featuring a cooperative strategy that integrates long-term frame band energy distribution selection with short-term frame subband gain optimization. At the long-term scale, a perceived annoyance prediction model quantitatively evaluates each critical band’s contribution to the target sound’s total annoyance, and an automatic selection mechanism chooses a controllable sound from candidate sounds that balances spectral similarity with a minimal sound pressure level (SPL) redundancy index. At the short-term scale, the simultaneous perturbation stochastic approximation gradient-descent (SPSA-GD) algorithm iteratively optimizes the gains of each critical band using annoyance as the cost function, enabling refined spectral reconstruction. Furthermore, to enhance system responsiveness, performance is analyzed and a variable step-size strategy is introduced, forming the VSS AAIM. The proposed methods are validated through single-channel simulations and experiments using range hood noise. Under hybrid operating condition, switching from strong turning to quick-fry mode, the VSS AAIM reduced actual annoyance by 1.01 and 1.16 compared to the fixed AIM, without a significant increase in SPL, achieving a perceptible improvement and demonstrating the method’s superiority.
KW - Adaptive control
KW - Audio injection method (AIM)
KW - Controllable sound optimization
UR - https://www.scopus.com/pages/publications/105054312170
U2 - 10.1016/j.apacoust.2026.111620
DO - 10.1016/j.apacoust.2026.111620
M3 - 文章
AN - SCOPUS:105054312170
SN - 0003-682X
VL - 257
JO - Applied Acoustics
JF - Applied Acoustics
M1 - 111620
ER -