plot(rate_K_sample_freqs_kHz,'+');
endfunction
-function [rate_K_surface rate_K_sample_freqs_kHz] = resample_const_rate_f_mel(model, K)
- rate_K_sample_freqs_kHz = mel_sample_freqs_kHz(K);
- rate_K_surface = resample_const_rate_f(model, rate_K_sample_freqs_kHz);
+function [rate_K_surface rate_K_sample_freqs_kHz] = resample_const_rate_f_mel(model, K, Fs=8000)
+ rate_K_sample_freqs_kHz = mel_sample_freqs_kHz(K, 100, 0.95*Fs/2);
+ rate_K_surface = resample_const_rate_f(model, rate_K_sample_freqs_kHz, Fs);
endfunction
% Resample Am from time-varying rate L=floor(pi/Wo) to fixed rate K. This can be viewed
% as a 3D surface with time, freq, and ampitude axis.
-function [rate_K_surface rate_K_sample_freqs_kHz] = resample_const_rate_f(model, rate_K_sample_freqs_kHz)
+function [rate_K_surface rate_K_sample_freqs_kHz] = resample_const_rate_f(model, rate_K_sample_freqs_kHz, Fs)
% convert rate L=pi/Wo amplitude samples to fixed rate K
- Fs = 8000; max_amp = 80;
+ max_amp = 160;
[frames col] = size(model);
K = length(rate_K_sample_freqs_kHz);
rate_K_surface = zeros(frames, K);
L = min([model(f,2) max_amp-1]);
Am = model(f,3:(L+2));
AmdB = 20*log10(Am);
- %pre = 10*log10((1:L)*Wo*4/(pi*0.3));
- %AmdB += pre;
- % clip between peak and peak -50dB, to reduce dynamic range
+ clip_en = 0;
+ if clip_en
+ % clip between peak and peak -50dB, to reduce dynamic range
- AmdB_peak = max(AmdB);
- AmdB(find(AmdB < (AmdB_peak-50))) = AmdB_peak-50;
-
- rate_L_sample_freqs_kHz = (1:L)*Wo*4/pi;
+ AmdB_peak = max(AmdB);
+ AmdB(find(AmdB < (AmdB_peak-50))) = AmdB_peak-50;
+ end
+
+ rate_L_sample_freqs_kHz = (1:L)*Wo*Fs/(2000*pi);
%rate_K_surface(f,:) = interp1(rate_L_sample_freqs_kHz, AmdB, rate_K_sample_freqs_kHz, "spline", "extrap");
rate_K_surface(f,:) = interp_para(rate_L_sample_freqs_kHz, AmdB, Fs/(2*1000), rate_K_sample_freqs_kHz);
% Take a rate K surface and convert back to time varying rate L
-function [model_ AmdB_] = resample_rate_L(model, rate_K_surface, rate_K_sample_freqs_kHz)
- max_amp = 80; Fs=8;
+function [model_ AmdB_] = resample_rate_L(model, rate_K_surface, rate_K_sample_freqs_kHz, Fs=8000)
+ max_amp = 1600;
[frames col] = size(model);
model_ = zeros(frames, max_amp+2);
for f=1:frames
Wo = model(f,1);
L = model(f,2);
- rate_L_sample_freqs_kHz = (1:L)*Wo*4/pi;
+ rate_L_sample_freqs_kHz = (1:L)*Wo*Fs/(2000*pi);
% back down to rate L
endfunction
-% Post Filter, has a big impact on speech quality after VQ. When used
+% PostFilter, has a big impact on speech quality after VQ. When used
% on a mean removed rate K vector, it raises formants, and supresses
% anti-formants. As it manipulates amplitudes, we normalise energy to
% prevent clipping or large level variations. pf_gain of 1.2 to 1.5
%model_ = experiment_filter(model);
%model_ = experiment_filter_dec_filter(model);
- [model_ surface] = experiment_rate_K_dct2(model, 0, 1, voicing);
- %[model_ surface] = experiment_mel_freq(model, 0, 1, voicing);
+ %[model_ surface] = experiment_rate_K_dct2(model, 0, 1, voicing);
+ [model_ surface] = experiment_mel_freq(model, 0, 1, voicing);
%model_ = experiment_smoothed(model, 0);
% break into 160ms blocks, 2D DCT, truncate, IDCT
- Tf = 0.01; % frame period in seconds
- Nt = 32; % number of 10ms frames blocks in time
- Nblocks = floor(frames/Nt);
+ Tf = 0.01; % frame period in seconds
+ Nt = 16; % number of 10ms frames blocks in time
+ dec = 2; % decimation factor
+ dist_dB = 2; % use enough coefficients to get this distortion ond DCT coeffs
+
+ Nblocks = floor(frames/(Nt*dec));
+ printf("frames: %d Nblocks: %d\n", frames, Nblocks);
unwrapped_dcts = zeros(Nblocks,Nt*K);
rate_K_surface_ = zeros(frames, K);
% create map on the fly from train database
- asurf = load("all_surf.txt");
+ asurf = load("all_surf.txt"); [nr nc] = size(asurf);
+ asurf = asurf(1:dec:nr,:);
[map rms_map mx mx_ind unwrapped_dcts] = create_map_rms(asurf, Nt, K);
%map = create_zigzag_map(Nt,K);
end
for n=1:Nblocks
- st = (n-1)*Nt+1; en = st + Nt - 1;
- D = dct2(rate_K_surface(st:en,:));
+ st = (n-1)*dec*Nt+1; en = st + dec*Nt - 1;
+ %printf("st: %d en: %d\n", st, en);
+ D = dct2(rate_K_surface(st:dec:en,:));
% move over surface and work out quantiser
% quantise, replace on map
qn = 0;
adct2_sd = mean(std(D-E));
- while adct2_sd > 3
+ while adct2_sd > dist_dB
qn++;
E(rmap(qn), cmap(qn)) = 1*round(D(rmap(qn), cmap(qn))/1);
adct2_sd = mean(std(D-E));
end
sumnz(n) = qn;
- rate_K_surface_(st:en,:) = idct2(E);
+ % note neat trick to interpolate to 10ms frames despite dec
+
+ #{
+ energy = sum(sum(E));
+ Edc = E(1,1);
+ E = E*1.2;
+ E(1,1) = Edc/1.2;
+ #}
+ %E *= energy/sum(sum(E));
+ rate_K_surface_(st:en,:) = idct2([sqrt(dec)*E; zeros(Nt*(dec-1), K)]);
+
dct2_sd(n) = mean(std(D-E));
end
- figure(3); clf; mesh(mapped);
+ % figure(3); clf; mesh(mapped);
figure(4); clf; plot(sumnz); hold on; plot([1 length(sumnz)],[mean(sumnz) mean(sumnz)]); hold off; title('Non Zero');
figure(5); clf; plot(dct2_sd); title('DCT SD');
printf("average dct spectral distortion: %3.2f dB\n", mean(dct2_sd));
printf("mean number of coeffs/DCT: %3.2f/%d\n", mean(sumnz), Nt*K);
- printf("coeffs/second: %3.2f\n", mean(sumnz)/(Nt*Tf));
- printf("bits/s: %3.2f\n", 2.9*mean(sumnz)/(Nt*Tf));
+ printf("coeffs/second: %3.2f\n", mean(sumnz)/(Nt*Tf*dec));
+ printf("bits/s: %3.2f\n", 2.9*mean(sumnz)/(Nt*Tf*dec));
+
+ % optional 700C style post filter
+
+ post_filter_en = 0;
+ if post_filter_en
+ for f=1:Nt*Nblocks
+ mn = mean(rate_K_surface_(f,:));
+ rate_K_surface_no_mean_(f,:) = rate_K_surface_(f,:) - mn;
+ rate_K_surface_(f,:) = mn + post_filter(rate_K_surface_no_mean_(f,:), rate_K_sample_freqs_kHz);
+ end
+ end
% prevent /0 errors at end of run
- rate_K_surface_(Nt*Nblocks+1:frames,:) = rate_K_surface(Nt*Nblocks+1:frames,:);
+ rate_K_surface_(dec*Nt*Nblocks+1:frames,:) = rate_K_surface(dec*Nt*Nblocks+1:frames,:);
model_ = resample_rate_L(model, rate_K_surface_, rate_K_sample_freqs_kHz);
- dist = std((rate_K_surface_ - rate_K_surface)');
+ dist = std((rate_K_surface_(1:dec:frames,:) - rate_K_surface(1:dec:frames,:))');
figure(1); clf; plot(dist); title('Rate K SD');
printf("Rate K spectral distortion mean: %3.2f dB var: %3.2f\n", mean(dist), var(dist));
endfunction
function [model_ rate_K_surface] = experiment_mel_freq(model, vq_en=0, plots=1, voicing)
[frames nc] = size(model);
- K = 20; Fs = 8000; correct_rate_K_en = 1;
+ K = 20; Fs = 16000; correct_rate_K_en = 0;
for f=1:frames
Wo = model(f,1);
L = model(f,2);
Am = model(f,3:(L+2));
AmdB = 20*log10(Am);
- Am_freqs_kHz = (1:L)*Wo*4/pi;
- [rate_K_vec rate_K_sample_freqs_kHz] = resample_const_rate_f_mel(model(f,:), K);
+ Am_freqs_kHz = (1:L)*Wo*Fs/(2000*pi);
+ [rate_K_vec rate_K_sample_freqs_kHz] = resample_const_rate_f_mel(model(f,:), K, Fs);
if correct_rate_K_en
- [tmp_ AmdB_] = resample_rate_L(model(f,:), rate_K_vec, rate_K_sample_freqs_kHz);
+ [tmp_ AmdB_] = resample_rate_L(model(f,:), rate_K_vec, rate_K_sample_freqs_kHz, Fs);
[rate_K_vec_corrected orig_error error nasty_error_log nasty_error_m_log] = correct_rate_K_vec(rate_K_vec, rate_K_sample_freqs_kHz, AmdB, AmdB_, K, Wo, L, Fs);
rate_K_surface(f,:) = rate_K_vec_corrected;
else
end
end
- model_ = resample_rate_L(model, rate_K_surface, rate_K_sample_freqs_kHz);
-
+ model_ = resample_rate_L(model, rate_K_surface, rate_K_sample_freqs_kHz, Fs);
+
endfunction
projects / freetel-svn-tracking.git / commitdiff