Quasi-Newton simultaneous perturbation stochastic approximation algorithm for broadband active noise control (L).

Area of Science:

• Acoustics and Signal Processing
• Control Systems Engineering
• Computational Mathematics

Background:

• Active noise control (ANC) is crucial for reducing unwanted sound.
• Existing methods, like SPSA, face challenges in efficient broadband noise reduction.
• Gradient estimation and filter updates are key to ANC performance.

Purpose of the Study:

• To develop an advanced ANC algorithm for enhanced broadband noise reduction.
• To integrate simultaneous perturbation stochastic approximation (SPSA) with the quasi-Newton method.
• To improve convergence speed and reduce steady-state error in ANC systems.

Main Methods:

• Proposed an ANC algorithm integrating SPSA and quasi-Newton methods.
• Developed efficient gradient estimation and simultaneous update of control filter and inverse Hessian matrix.
• Validated the algorithm using in-vehicle noise data.

Main Results:

• The proposed ANC algorithm demonstrated faster convergence compared to existing SPSA methods.
• Achieved a lower steady-state mean square error.
• Showcased effectiveness in real-world in-vehicle noise scenarios with minimal computational overhead.

Conclusions:

• The integrated SPSA and quasi-Newton ANC algorithm significantly enhances broadband noise reduction.
• The method offers improved performance metrics (convergence, error) over traditional SPSA.
• This approach presents a computationally efficient solution for practical ANC applications.