Adjusting Parameters for G.729 Voice Activity Detection Algorithm
1 回表示 (過去 30 日間)
古いコメントを表示
I am currently attempting to use the MATLAB implementation of the G.729 VAD algorithm (included with my version of MATLAB 2015b). In it's default state it has too many false positive decision for my application (voice in the presence of background noise). So I am trying to figure out which parameters I can adjust to achieve a less sensitive response.
I have searched for an explanation of the parameters used in G.729 VAD but have not found anything that is useful on this topic. It is possible I am searching incorrectly.
The parameters as they come in the MATLAB DSP System Toolbox function are shown below
%%Algorithm Constants Initialization
VAD_cst_param.Fs = single(8000); % Sampling Frequency
VAD_cst_param.M = single(10); % Order of LP filter
VAD_cst_param.NP = single(12); % Increased LPC order
VAD_cst_param.No = single(32); % Number of frames for long-term minimum energy calculation
VAD_cst_param.Ni = single(32); % Number of frames for initialization of running averages
VAD_cst_param.INIT_COUNT = single(20);
% High Pass Filter that is used to preprocess the signal applied to the VAD
VAD_cst_param.HPF_sos = single([0.92727435, -1.8544941, 0.92727435, 1, -1.9059465, 0.91140240]);
VAD_cst_param.L_WINDOW = single(240); % Window size in LP analysis
VAD_cst_param.L_NEXT = single(40); % Lookahead in LP analysis
VAD_cst_param.L_FRAME = single(80); % Frame size
L1 = VAD_cst_param.L_NEXT;
L2 = VAD_cst_param.L_WINDOW - VAD_cst_param.L_NEXT;
VAD_cst_param.hamwindow = single([0.54 - 0.46*cos(2*pi*(0:L2-1)'/(2*L2-1));
cos(2*pi*(0:L1-1)'/(4*L1-1))]);
% LP analysis, lag window applied to autocorrelation coefficients
VAD_cst_param.lagwindow = single([1.0001; exp(-1/2 * ((2 * pi * 60 / VAD_cst_param.Fs) * (1:VAD_cst_param.NP)').^2)]);
% Correlation for a lowpass filter (3 dB point on the power spectrum is
% at about 2 kHz)
VAD_cst_param.lbf_corr = single([0.24017939691329, 0.21398822343783, 0.14767692339633, ...
0.07018811903116, 0.00980856433051,-0.02015934721195, ...
-0.02388269958005,-0.01480076155002,-0.00503292155509, ...
0.00012141366508, 0.00119354245231, 0.00065908718613, ...
0.00015015782285]');
0 件のコメント
回答 (0 件)
参考
カテゴリ
Help Center および File Exchange で Linear Prediction についてさらに検索
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!