mel vq quantiser working OK at 18 bits/fr, added some sample command lines

git-svn-id: https://svn.code.sf.net/p/freetel/code@2258 01035d8c-6547-0410-b346-abe4f91aad63

diff --git a/codec2-dev/octave/melvq.m b/codec2-dev/octave/melvq.m
index aab22033352f89deb5cd03f95910f2b6c688764b..7d5c47bd6da47d91144b54e8c35d83e81ea5cf75 100644 (file)
--- a/codec2-dev/octave/melvq.m
+++ b/codec2-dev/octave/melvq.m
@@ -12,6 +12,11 @@
 1;
 
 % train up multi-stage VQ
+% ~/codec2-dev/build_linux/src$ sox -r 8000 -s -2 ../../wav/all.wav -t raw -r 8000 -s -2 - sinc 300 sinc -2600 | ./c2sim - --lpc 6 --lpcpf --lspmel --dump all  -o - | play -t raw -r 8000 -s -2 - vol 3
+%
+% octave:> load ../build_linux/src/all_mel.txt
+% octave:> melvq; vq = trainvq(all_mel, 64, 3);
+% octave:> save vq
 
 function vq = trainvq(training_data, Nvec, stages)
 
@@ -51,14 +56,15 @@ endfunction
 
 % Search multi-stage VQ, retaining m best candidates at each stage
 
-function [res ind] = mbest(vqset, input_vecs, m)
+function [res output_vecs ind] = mbest(vqset, input_vecs, m)
 
   [Nvec order stages] = size(vqset);
   [Ninput tmp] = size(input_vecs);
 
-  res = [];   % residual error after VQ
-  ind = [];   % index of vqs
-
+  res = [];         % residual error after VQ
+  output_vecs = []; % quantised ouput vectors
+  ind = [];         % index of vqs
+  
   for i=1:Ninput
   
     % first stage, find mbest candidates
@@ -105,26 +111,54 @@ function [res ind] = mbest(vqset, input_vecs, m)
 
     % final residual
     target(1,:) = input_vecs(i,:);
-    for v=1:s
+    out = zeros(1,order);
+    for v=1:stages
       target(1,:) -= vqset(cand_list(1,v+1),:,v);
+      out += vqset(cand_list(1,v+1),:,v);
     end
-    res = [res; target(1,:)];
-    ind = [ind; cand_list(1,2:1+stages)];
+    res  = [res; target(1,:)];
+    output_vecs  = [output_vecs; out];
+    ind  = [ind; cand_list(1,2:1+stages)];
   end
 
 endfunction
 
-more off;
-load "../build_linux/src/all_mel.txt"
-load vq;
-[res1 ind] = mbest(vq, all_mel(1,:),3);
-mean(std(res1))
-
-% save text file of "vq quantised mels"
-% load back into c2sim at run time
-% train on continuous mels
-% sorting/stability
-% see how it sounds
-% Goal is to get VQ sounding OK, similar to UQ at 20 or 40ms dec,
-% sig better than current 700
-% analysis of data, outliers, different training
+
+% Quantises a set of msl-lsps and saves back to disk so they can be read in by c2sim
+% assumes we have a vq saved to disk called vq
+%
+% ~/codec2-dev/build_linux/src$ sox -r 8000 -s -2 ../../wav/vk5qi.wav -t raw -r 8000 -s -2 - sinc 300 sinc -2600 | ./c2sim - --lpc 6 --lpcpf --lspmel --dump vk5qi  -o - | play -t raw -r 8000 -s -2 - vol 3
+%
+% octave:> test_run("vk5qi")
+%
+% ~/codec2-dev/build_linux/src$ sox -r 8000 -s -2 ../../wav/vk5qi.wav -t raw -r 8000 -s -2 - sinc 300 sinc -2600 | ./c2sim - --lpc 6 --lpcpf --phase0 --dec 4 --postfilter --lspmel --lspmelread ../../octave/vk5qi_mel_.out -o - | play -t raw -r 8000 -s -2 - vol 3
+
+function ind = test_run(samplename)
+
+  more off;
+  input_vecs_name = sprintf("../build_linux/src/%s_mel.txt", samplename);
+  input_vecs_name
+  mel = load(input_vecs_name);
+  load vq;
+  [res mel_ ind] = mbest(vq, mel, 5);
+  mean(std(res))
+
+  output_vecs_name = sprintf("%s_mel_.out", samplename);
+  fmel_ = fopen(output_vecs_name,"wb"); 
+  [r c] = size(mel_);
+  for i=1:r
+    fwrite(fmel_, mel_(i,:), "float32"); 
+  end
+  fclose(fmel_);
+end
+
+ind = test_run("all")
+
+% [X] save text file of "vq quantised mels"
+% [X] load back into c2sim at run time
+% [X] train on continuous mels
+% [X] sorting/stability
+% [X] see how it sounds
+% [X] Goal is to get VQ sounding OK, similar to UQ at 20 or 40ms dec,
+% [X] sig better than current 700
+% [X] check all indexes used with hist

diff --git a/codec2-dev/src/c2sim.c b/codec2-dev/src/c2sim.c
index 89afbd04bdd5f3363d229bccfb5faedd995ab8d0..814a03060be14456295078267b98929ea219d289 100644 (file)
--- a/codec2-dev/src/c2sim.c
+++ b/codec2-dev/src/c2sim.c
@@ -104,9 +104,10 @@ int main(int argc, char *argv[])
 
     int   postfilt;
 
-    int   hand_voicing = 0, phaseexp = 0, ampexp = 0, hi = 0, simlpcpf = 0;
+    int   hand_voicing = 0, phaseexp = 0, ampexp = 0, hi = 0, simlpcpf = 0, lspmelread = 0;
     int   lpcpf = 0;
     FILE *fvoicing = 0;
+    FILE *flspmel = 0;
 
     MODEL prev_model;
     int dec;
@@ -141,6 +142,7 @@ int main(int argc, char *argv[])
         { "lpc", required_argument, &lpc_model, 1 },
         { "lspjnd", no_argument, &lspjnd, 1 },
         { "lspmel", no_argument, &lspmel, 1 },
+        { "lspmelread", required_argument, &lspmelread, 1 },
         { "lsp", no_argument, &lsp, 1 },
         { "lspd", no_argument, &lspd, 1 },
         { "lspvq", no_argument, &lspvq, 1 },
@@ -255,6 +257,12 @@ int main(int argc, char *argv[])
                        optarg, strerror(errno));
                     exit(1);
                 }
+            } else if(strcmp(long_options[option_index].name, "lspmelread") == 0) {
+               if ((flspmel = fopen(optarg,"rb")) == NULL) {
+                   fprintf(stderr, "Error opening float lspmel file: %s: %s.\n",
+                       optarg, strerror(errno));
+                    exit(1);
+                }
            } else if(strcmp(long_options[option_index].name, "dump_pitch_e") == 0) {
                if ((fjvm = fopen(optarg,"wt")) == NULL) {
                    fprintf(stderr, "Error opening pitch & energy dump file: %s: %s.\n",
@@ -624,41 +632,35 @@ int main(int argc, char *argv[])
                    mel[i] = floor(2595.0*log10(1.0 + f/700.0) + 0.5);
                }
 
-                /* round to the nearest x hz */
-                #define MEL_ROUND 50
-               for(i=0; i<order; i++) {
-                    float x = floor(mel[i]/MEL_ROUND+0.5)*MEL_ROUND;
-                    mel[i] = x;
-                    //printf("mel[%d] = %f x: %f\n", i, mel[i], x);
-                }
-
-               for(i=1; i<order; i++) {
-                   if (mel[i] == mel[i-1]) {
-                       mel[i]+=MEL_ROUND/2;
-                       mel[i-1]-=MEL_ROUND/2;
-                        i = 1;
-                    }
-               }
+                #ifdef DUMP
+                dump_mel(mel, order);
+                #endif
 
                encode_mels_scalar(mel_indexes, mel, 6);
                 #ifdef DUMP
                 dump_mel_indexes(mel_indexes, 6);
                 #endif
-               decode_mels_scalar(mel, mel_indexes, 6);
+               //decode_mels_scalar(mel, mel_indexes, 6);
                 
+                /* read in VQed lsp-mels from octave/melvq.m */
+
+                if (lspmelread) {
+                    int ret = fread(mel, sizeof(float), order, flspmel);
+                    assert(ret == order);
+                }
+                                
+                /* ensure no unstable filters after quantisation */
+
+                #define MEL_ROUND 50
                for(i=1; i<order; i++) {
-                   if (mel[i] == mel[i-1]) {
+                   if (mel[i] <= mel[i-1]) {
                        mel[i]+=MEL_ROUND/2;
                        mel[i-1]-=MEL_ROUND/2;
                         i = 1;
                     }
                }
-
-                #ifdef DUMP
-                dump_mel(mel, order);
-                #endif
-
-               for(i=0; i<LPC_ORD; i++) {
+                
+               for(i=0; i<order; i++) {
                    f_ = 700.0*( pow(10.0, mel[i]/2595.0) - 1.0);
                    lsps_[i] = f_*(PI/4000.0);
                }

diff --git a/codec2-dev/wav/all.wav b/codec2-dev/wav/all.wav
new file mode 100644 (file)
index 0000000..ba37838
Binary files /dev/null and b/codec2-dev/wav/all.wav differ