howling

https://theses.eurasip.org/media/theses/documents/van-waterschoot-toon-design-and-evaluation-of-digital-signal-processing-algorithms-for-acoustic-feedback-and-echo-cancellation.pdf

https://ggwujun.github.io/blog-front/%E6%90%9E%E5%AE%9A%E9%9F%B3%E9%A2%91%E6%8A%80%E6%9C%AF/

https://github.com/audiolabs/rir-generator

https://github.com/audiolabs/anechoic-noise

https://github.com/audiolabs/torch-pesq

https://github.com/audiolabs/SC-Wind-Noise-Generator

https://github.com/audiolabs/anf-generator

https://github.com/audiolabs/blind-multi-room-model

https://github.com/audiolabs/PESQ

https://github.com/audiolabs/webMUSHRA

https://github.com/Ryuk17/Aura/tree/master

https://blog.csdn.net/weixin_42141751

https://github.com/aflyingwolf

http://www.kaldi-asr.org/

https://htk.eng.cam.ac.uk/

https://www.cnblogs.com/lulululuyan/p/18141262

https://zhuanlan.zhihu.com/p/679689442

https://www.aigei.com/sound/class/xiao_jiao/

https://doc.zh-jieli.com/AC104/zh-cn/master/sound_effect/notch_howling/notch_howling.html

Howling detection based on LMS algorithm

https://blog.csdn.net/hmq252198885/article/details/144721693

% LMS算法的MATLAB实现
function [w, e] = LMS(d, x, mu, M)
% 输入:
% d – 期望信号
% x – 输入信号
% mu – 步长
% M – 滤波器阶数
% 输出:
% w – 滤波器权重
% e – 误差信号
L = length(d);
w = zeros(M, 1);
e = zeros(L, 1);
for n = M:L
x_n = x(n:-1:n-M+1); % 输入信号的当前样本和前M-1个样本
y_n = w’ * x_n; % 滤波器输出
e(n) = d(n) – y_n; % 误差信号
w = w + mu * e(n) * x_n; % 权重更新
end
end

% NLMS算法的MATLAB实现
function [w, e] = NLMS(d, x, mu, M)
% 输入:
% d – 期望信号
% x – 输入信号
% mu – 步长
% M – 滤波器阶数
% 输出:
% w – 滤波器权重
% e – 误差信号
L = length(d);
w = zeros(M, 1);
e = zeros(L, 1);
for n = M:L
x_n = x(n:-1:n-M+1); % 输入信号的当前样本和前M-1个样本
y_n = w’ * x_n; % 滤波器输出
e(n) = d(n) – y_n; % 误差信号
P = x_n’ * x_n; % 输入信号的功率
w = w + mu * e(n) * x_n / (P + 1e-6); % 权重更新，加入小量避免除零
end
end