A neural reference projection-based method for multi-reference active noise control (L)

Area of Science:

• Engineering
• Signal Processing
• Machine Learning

Background:

• Active noise control (ANC) systems require efficient processing of multichannel reference signals.
• Conventional methods for reference signal compression can limit performance in complex acoustic environments.

Purpose of the Study:

• To propose a novel neural reference projection-based method for compressing multichannel reference signals in feedforward ANC.
• To integrate this method with the filtered-x affine projection algorithm to enhance ANC performance.

Main Methods:

• A neural reference projection technique was developed for signal compression.
• The proposed method was combined with the filtered-x affine projection algorithm.
• Simulations were conducted using a real-vehicle road noise dataset.

Main Results:

• The neural reference projection-based filtered-x affine projection algorithm demonstrated faster convergence compared to conventional methods.
• Higher broadband noise attenuation was achieved in multi-reference ANC scenarios.
• The proposed method effectively improved both convergence speed and noise reduction performance.

Conclusions:

• The neural reference projection method offers a significant advancement in feedforward ANC.
• This technique provides superior performance for real-world applications like in-vehicle noise reduction.
• The integration with filtered-x affine projection enhances the efficiency and effectiveness of ANC systems.