matlab-基于LCMV-GSC算法的语音增加仿真-源码

function [ scores ] = pesq( ref_wav, deg_wav ) if nargin ==0, fprintf('Usage: pesq( ref_wav, deg_wav )\n'); fprintf(' ref_wav = reference input filename\n'); fprintf(' deg_wav = degraded output filename\n\n'); fprintf('For more help, type: help pesq\n\n'); return; elseif nargin > 2, error('%s.m: Incorrect number of input arguments.\nFor usage help type: help %s',mfilename,mfilename); end if ~isstr( ref_wav ), error( '%s.m: First input argumnet has to be a reference wav filename as string.\nFor usage help type: help %s',mfilename,mfilename); end; if ~isstr( deg_wav ), error( '%s.m: Second input argumnet has to be a processed wav filename as string.\nFor usage help type: help %s',mfilename,mfilename); end; if ~exist( ref_wav, 'file' ), error( '%s.m: Reference wav file: %s not found.',mfilename,ref_wav); end; if ~exist( deg_wav, 'file' ), error( '%s.m: Processed wav file: %s not found.',mfilename,deg_wav); end; [ ref_data, ref_sampling_rate ] = audioread( ref_wav ); [ deg_data, deg_sampling_rate ] = audioread( deg_wav ); if ref_sampling_rate ~= deg_sampling_rate, error( '%s.m: Sampling rate mismatch.\nThe sampling rate of the reference wav file (%i Hz) has to match sampling rate of the degraded wav file (%i Hz).',mfilename,ref_sampling_rate,deg_sampling_rate); else, sampling_rate = ref_sampling_rate; end; if sampling_rate==8E3, mode='narrowband'; elseif sampling_rate==16E3, mode='wideband'; else, error( '%s.m: Unsupported sampling rate (%i Hz).\nOnly sampling rates of 8000 Hz (for narrowband assessment)\nand 16000 Hz (for wideband assessment) are supported.',mfilename,sampling_rate); end clearvars -global Downsample DATAPADDING_MSECS SEARCHBUFFER Fs WHOLE_SIGNAL Align_Nfft Window global Downsample DATAPADDING_MSECS SEARCHBUFFER Fs WHOLE_SIGNAL global Align_Nfft Window setup_global( sampling_rate ); TWOPI= 6.28318530717959; for count = 0: Align_Nfft- 1 Window(1+ count) = 0.5 * (1.0 - cos((TWOPI * count) / Align_Nfft)); end ref_data= ref_data(:).'; ref_data= ref_data* 32768; ref_Nsamples= length( ref_data)+ 2* SEARCHBUFFER* Downsample; ref_data= [zeros( 1, SEARCHBUFFER* Downsample), ref_data, ... zeros( 1, DATAPADDING_MSECS* (Fs/ 1000)+ SEARCHBUFFER* Downsample)]; deg_data= deg_data(:).'; deg_data= deg_data* 32768; deg_Nsamples= length( deg_data)+ 2* SEARCHBUFFER* Downsample; deg_data= [zeros( 1, SEARCHBUFFER* Downsample), deg_data, ... zeros( 1, DATAPADDING_MSECS* (Fs/ 1000)+ SEARCHBUFFER* Downsample)]; maxNsamples= max( ref_Nsamples, deg_Nsamples); ref_data= fix_power_level( ref_data, ref_Nsamples, maxNsamples); deg_data= fix_power_level( deg_data, deg_Nsamples, maxNsamples); % KKW --------- switch lower( mode ) case { [], '', 'nb', '+nb', 'narrowband', '+narrowband' } standard_IRS_filter_dB= [0, -200; 50, -40; 100, -20; 125, -12; 160, -6; 200, 0;... 250, 4; 300, 6; 350, 8; 400, 10; 500, 11; 600, 12; 700, 12; 800, 12;... 1000, 12; 1300, 12; 1600, 12; 2000, 12; 2500, 12; 3000, 12; 3250, 12;... 3500, 4; 4000, -200; 5000, -200; 6300, -200; 8000, -200]; ref_data= apply_filter( ref_data, ref_Nsamples, standard_IRS_filter_dB); deg_data= apply_filter( deg_data, deg_Nsamples, standard_IRS_filter_dB); case { 'wb', '+wb', 'wideband', '+wideband' } ref_data = apply_filters_WB( ref_data, ref_Nsamples ); deg_data = apply_filters_WB( deg_data, deg_Nsamples ); otherwise error( sprintf('Mode: "%s" is unsupported.', mode) ); end % ------------- % % fid= fopen( 'log_mat_ref.txt', 'wt'); % fprintf( fid, '%f\n', ref_data); % fclose( fid); % % fid= fopen( 'log_mat_deg.txt', 'wt'); % fprintf( fid, '%f\n', deg_data); % fclose( fid); % % to save time, read from data file ======== % fid= fopen( 'log_mat_ref.txt', 'rt'); % ref_data= fscanf( fid, '%f\n'); % ref_data= ref_data'; % fclose( fid); % ref_Nsamples= length( ref_data)- DATAPADDING_MSECS* (Fs/ 1000); % % fid= fopen( 'log_mat_deg.txt', 'rt'); % deg_data= fscanf( fid, '%f\n'); % deg_data= deg_data'; % fclose( fid); % deg_Nsamples= length( deg_data)- DATAPADDING_MSECS* (Fs/ 1000); % % the above part will be commented after debugging ======== % for later use in psychoacoustical model model_ref= ref_data; model_deg= deg_data; [ref_data, deg_data]= input_filter( ref_data, ref_Nsamples, deg_data, ... deg_Nsamples); % fid= fopen( 'log_mat_ref_tovad.txt', 'wt'); % fprintf( fid, '%f\n', ref_data); % fclose( fid); % % fid= fopen( 'log_mat_deg_tovad.txt', 'wt'); % fprintf( fid, '%f\n', deg_data); % fclose( fid); [ref_VAD, ref_logVAD]= apply_VAD( ref_data, ref_Nsamples); [deg_VAD, deg_logVAD]= apply_VAD( deg_data, deg_Nsamples); % subplot( 2, 2, 1); plot( ref_VAD); title( 'ref\_VAD'); % subplot( 2, 2, 2); plot( ref_logVAD); title( 'ref\_logVAD'); % % subplot( 2, 2, 3); plot( deg_VAD); title( 'deg\_VAD'); % subplot( 2, 2, 4); plot( deg_logVAD); title( 'deg\_logVAD'); % % fid= fopen( 'mat_ref_vad.txt', 'wt'); % fprintf( fid, '%f\n', ref_VAD); % fclose( fid); % % fid= fopen( 'mat_ref_logvad.txt', 'wt'); % fprintf( fid, '%f\n', ref_logVAD); % fclose( fid); % % fid= fopen( 'mat_deg_vad.txt', 'wt'); % fprintf( fid, '%f\n', deg_VAD); % fclose( fid); % % fid= fopen( 'mat_deg_logvad.txt', 'wt'); % fprintf( fid, '%f\n', deg_logVAD); % fclose( fid); % crude_align (ref_logVAD, ref_Nsamples, deg_logVAD, deg_Nsamples,... WHOLE_SIGNAL); utterance_locate (ref_data, ref_Nsamples, ref_VAD, ref_logVAD,... deg_data, deg_Nsamples, deg_VAD, deg_logVAD); ref_data= model_ref; deg_data= model_deg; % make ref_data and deg_data equal length if (ref_Nsamples< deg_Nsamples) newlen= deg_Nsamples+ DATAPADDING_MSECS* (Fs/ 1000); ref_data( newlen)= 0; elseif (ref_Nsamples> deg_Nsamples) newlen= ref_Nsamples+ DATAPADDING_MSECS* (Fs/ 1000); deg_data( newlen)= 0; end pesq_mos= pesq_psychoacoustic_model (ref_data, ref_Nsamples, deg_data, ... deg_Nsamples ); % KKW --------- switch lower( mode ) case { [], '', 'nb', '+nb', 'narrowband', '+narrowband' } % NB: P.862.1->P.800.1 (PESQ_MOS->MOS_LQO) mos_lqo = 0.999 + ( 4.999-0.999 ) ./ ( 1+exp(-1.4945*pesq_mos+4.6607) ); scores = [ pesq_mos, mos_lqo ]; case { 'wb', '+wb', 'wideband', '+wideband' } % WB: P.862.2->P.800.1 (PESQ_MOS->MOS_LQO) mos_lqo = 0.999 + ( 4.999-0.999 ) ./ ( 1+exp(-1.3669*pesq_mos+3.8224) ); scores = [ mos_lqo ]; otherwise error( sprintf('Mode: "%s" is unsupported.', mode) ); end % ------------- %fprintf( '\tPrediction PESQ_MOS = %4.3f\n', pesq_mos ); clearvars -global Downsample DATAPADDING_MSECS SEARCHBUFFER Fs WHOLE_SIGNAL Align_Nfft Window function align_filtered= apply_filter( data, data_Nsamples, align_filter_dB) global Downsample DATAPADDING_MSECS SEARCHBUFFER Fs align_filtered= data; n= data_Nsamples- 2* SEARCHBUFFER* Downsample+ DATAPADDING_MSECS* (Fs/ 1000); % now find the next power of 2 which is greater or equal to n pow_of_2= 2^ (ceil( log2( n))); [number_of_points, trivial]= size( align_filter_dB); overallGainFilter= interp1( align_filter_dB( :, 1), align_filter_dB( :, 2), ... 1000); x= zeros( 1, pow_of_2); x( 1: n)= data( SEARCHBUFFER* Downsa