Thanks for your sharing for ANC.
When I use a headphone Pri and Sec Path, I found GFANC can't work well.
In GFANC Paper, The Pri and Sec Path are bandpass FIR filter. I try to use our measure path, the mag response are below.

Pri Path:

Sec Path:

I put 20~7980Hz white noise signal into train_fxlms_algorithm to create Pretrained_Sub_Control_filters. The fxlms output result is below.

Using the above Pretrained_Sub_Control_filters to train. The Accuracy is not good.

Epoch 37
Learning rate: 7.8125e-05
Training Loss: 0.12145824613980949 Training Accuracy: 0.5167499999999995
Validation Loss : 0.12397415339946746 Validation Accuracy : 0.5225000000000001

And the GFANC result for Aircraft nosie is also not good.

I think there is a lot of nolinear in our pri and sec path and the fxlms can't get a ideal FIR filter.

Hi,

I have read your work, and it is very interesting. Thank you for sharing this with the community. I would like to use your algorithm with other control paths, but the control is not applied, so I would like to retrain the model with my data. Therefore, I wanted to know if it is possible to share the training code of your model as well.

Kind regards,

Does simply modify "Pretrained_Sub_Control_filters.mat" enough ?

Hi there,

First of all, I want to express my appreciation for your work on SFANC-FxNLMS and GFANC. It's inspiring to see someone working on open-source ANC systems, and your assistance has been incredibly helpful to me.

However, I have encountered a question that I am struggling to understand. Both SFANC and GFANC utilize the Control_filter_selection() function to generate the filters that SFANC and GFANC required (e.g., 10 filters for a 10-second duration). But, during the actual SFANC and GFANC stages, only 9 filters are utilized. This discrepancy arises because the first-second noise uses the filter that are entirely composed of zeros, and the rest of "i" second noise uses the "i - 1"th filter, and nobody utilize the last filter. I am puzzled as to why the "i" second noise cannot use the "i"th filter directly.

In the GFANC paper, it is mentioned that:
"However, GFANC and SFANC cannot handle the first-second noise because they update the control filter coefficients for the next second based on the first-second noise."

However, in the program, I haven't come across any code segment that explicitly updates the control filter coefficients based on the first-second noise. Instead, the control filter coefficients appear to be switched to self.Filter_vector[j] at the end of each second.

Could you please help clarify this discrepancy for me? I am eager to gain a better understanding of how the control filter coefficients are updated in SFANC-FxNLMS and GFANC. Thank you in advance for your assistance.