Experimental index optimised direct VQ

.github/workflows/cmake.yml

@@ -0,0 +1,39 @@
+name: CMake
+
+on: [pull_request]
+
+env:
+  # Customize the CMake build type here (Release, Debug, RelWithDebInfo, etc.)
+  BUILD_TYPE: Debug
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v2
+
+    - name: Install packages
+      shell: bash
+      run: |
+           sudo apt-get update
+           sudo apt-get install octave octave-common octave-signal gnuplot sox p7zip-full
+
+    - name: Build codec2
+      shell: bash
+      run: |
+           git clone https://github.com/drowe67/codec2.git 
+           cd codec2 && mkdir -p build_linux && cd build_linux 
+           cmake ..
+           make
+           pwd
+
+    - name: Build LPCNet
+      shell: bash
+      run: |
+           mkdir -p build_linux && cd build_linux 
+           cmake -DCODEC2_BUILD_DIR=$GITHUB_WORKSPACE/codec2/build_linux ..
+           make
+           # simple test to make sure the code runs
+           cd src && sox ../../wav/wia.wav -t raw -r 16000 - | ./lpcnet_enc -s | ./lpcnet_dec -s > /dev/null
+

src/CMakeLists.txt

@@ -12,10 +12,10 @@ set(lpcnet_freedv_srcs
     lpcnet_dump.c
     4stage_pred_vq.c
     4stage_direct_split_vq.c
+    4stage_direct_split_indopt_vq.c
     lpcnet.c
     lpcnet_freedv.c
     nnet.c
-    nnet_rw.c
     ${lpcnet_SOURCE_DIR}/nnet_data.c
 )
 
@@ -48,7 +48,7 @@ export(TARGETS lpcnetfreedv
 add_executable(dump_data dump_data.c)
 target_link_libraries(dump_data lpcnetfreedv m codec2)
 
-add_executable(test_lpcnet test_lpcnet.c)
+add_executable(test_lpcnet test_lpcnet.c nnet_rw.c)
 target_link_libraries(test_lpcnet lpcnetfreedv m codec2)
 
 if(SSE OR AVX OR AVX2 OR CMAKE_SYSTEM_PROCESSOR STREQUAL "aarch64")
@@ -91,13 +91,13 @@ target_link_libraries(quant_dec lpcnetfreedv m codec2)
 add_executable(lpcnet_enc lpcnet_enc.c)
 target_link_libraries(lpcnet_enc lpcnetfreedv m codec2)
 
-add_executable(lpcnet_dec lpcnet_dec.c)
+add_executable(lpcnet_dec lpcnet_dec.c nnet_rw.c)
 target_link_libraries(lpcnet_dec lpcnetfreedv m codec2)
 
 add_executable(idct idct.c)
 target_link_libraries(idct lpcnetfreedv m codec2)
 
-add_executable(nnet2f32 nnet2f32.c)
+add_executable(nnet2f32 nnet2f32.c nnet_rw.c)
 target_link_libraries(nnet2f32 lpcnetfreedv m)
 
 add_executable(sw2packedulaw sw2packedulaw.c)

src/lpcnet.c

@@ -133,10 +133,7 @@ void lpcnet_destroy(LPCNetState *lpcnet)
 
 void lpcnet_open_test_file(LPCNetState *lpcnet, char file_name[]) {
     lpcnet->ftest = fopen(file_name, "wb");
-    if (lpcnet->ftest == NULL) {
-        fprintf(stderr, "Error opening LPCNet test file: %s\n", file_name);
-        exit(1);
-    }
+    assert(lpcnet->ftest != NULL);
 }
 
 void lpcnet_set_preemph(LPCNetState *lpcnet, float preemph) {

src/lpcnet_dec.c

@@ -61,8 +61,9 @@ int main(int argc, char **argv) {
     int   *m = pred_m;
     float *vq = pred_vq;
     int   logmag = 0;
-    int   direct_split = 0;
-
+    int   vq_type = LPCNET_PRED;
+    int   ber_st=0, ber_en=-1;
+
     fin = stdin;
     fout = stdout;
 
@@ -72,20 +73,23 @@ int main(int argc, char **argv) {
         {"infile",      required_argument, 0, 'i'},
         {"outfile",     required_argument, 0, 'u'},
         {"ber",         required_argument, 0, 'b'},
+        {"ber_st",      required_argument, 0, 'c'},
+        {"ber_en",      required_argument, 0, 'e'},
         {"decimate",    required_argument, 0, 'd'},
         {"nnet",        required_argument, 0, 'r'},
         {"numstages",   required_argument, 0, 'n'},
         {"pitchquant",  required_argument, 0, 'o'},
         {"pred",        required_argument, 0, 'p'},
-        {"directsplit", no_argument,       0, 's'},
+        {"split",       no_argument,       0, 's'},
+        {"indexopt",    no_argument,       0, 'x'},
         {"verbose",     no_argument,       0, 'v'},
         {0, 0, 0, 0}
     };
 
     int   c;
     int opt_index = 0;
 
-    while ((c = getopt_long (argc, argv, "b:d:n:o:p:svi:u:r:", long_options, &opt_index)) != -1) {
+    while ((c = getopt_long (argc, argv, "b:c:e:d:n:o:p:sxvi:u:r:", long_options, &opt_index)) != -1) {
         switch (c) {
  	case 'i':
             if ((fin = fopen(optarg, "rb")) == NULL) {
@@ -103,10 +107,16 @@ int main(int argc, char **argv) {
             ber = atof(optarg);
             fprintf(stderr, "BER = %f\n", ber);
             break;
+        case 'c':
+            ber_st = atoi(optarg);
+            break;
         case 'd':
             dec = atoi(optarg);
             fprintf(stderr, "dec = %d\n", dec);
             break;
+        case 'e':
+            ber_en = atoi(optarg);
+            break;
         case 'n':
             num_stages = atoi(optarg);
             fprintf(stderr, "%d VQ stages\n",  num_stages);
@@ -124,15 +134,23 @@ int main(int argc, char **argv) {
 	    nnet_read(optarg);
 	    break;            
         case 's':
-            direct_split = 1; m = direct_split_m; vq = direct_split_vq; pred = 0.0; logmag = 1; weight = 1.0;
-            fprintf(stderr, "split VQ\n");
+            vq_type = LPCNET_DIRECT_SPLIT;
+            m = direct_split_m; vq = direct_split_vq; pred = 0.0; logmag = 1; weight = 1.0;
+            fprintf(stderr, "direct split VQ\n");
+            break;
+        case 'x':
+            vq_type = LPCNET_DIRECT_SPLIT_INDEX_OPT;
+            m = direct_split_indopt_m; vq = direct_split_indopt_vq; pred = 0.0; logmag = 1; weight = 1.0;
+            fprintf(stderr, "index optimised direct split VQ\n");
             break;
         case 'v':
             lpcnet_verbose = 1;
             break;
         default:
             fprintf(stderr,"usage: %s [Options]:\n", argv[0]);
             fprintf(stderr,"  [-b --ber BER]\n");
+            fprintf(stderr,"  [--ber_st bit   Bit in frame where we start inserting errors (default 0)]\n");
+            fprintf(stderr,"  [--ber_en bit   Bit in frame just after we stop inserting errors (default 51)]\n");
             fprintf(stderr,"  [-d --decimation 1/2/3...]\n");
             fprintf(stderr,"  [-n --numstages]\n  [-o --pitchbits nBits]\n");
             fprintf(stderr,"  [-p --pred predCoff]\n");
@@ -142,7 +160,7 @@ int main(int argc, char **argv) {
         }
     }
 
-    LPCNetFreeDV *lf = lpcnet_freedv_create(direct_split);
+    LPCNetFreeDV *lf = lpcnet_freedv_create(vq_type);
     lpcnet_open_test_file(lf->net, "test_lpcnet_statesq.f32");
     LPCNET_QUANT *q = lf->q;
 
@@ -161,14 +179,15 @@ int main(int argc, char **argv) {
     char frame[q->bits_per_frame];
     int bits_read = 0;
     short pcm[lpcnet_samples_per_frame(lf)];
-
+    if (ber_en == -1) ber_en = q->bits_per_frame-1;
+
     do {
 
         bits_read = fread(frame, sizeof(char), q->bits_per_frame, fin);
-        nbits += bits_read;
+        nbits += ber_en - ber_st;
         if (ber != 0.0) {
             int i;
-            for(i=0; i<q->bits_per_frame; i++) {
+            for(i=ber_st; i<=ber_en; i++) {
                 float r = (float)rand()/RAND_MAX;
                 if (r < ber) {
                     frame[i] = (frame[i] ^ 1) & 0x1;
@@ -190,6 +209,7 @@ int main(int argc, char **argv) {
     lpcnet_freedv_destroy(lf);
 
     if (ber != 0.0)
-        fprintf(stderr, "nbits: %d nerr: %d BER: %4.3f\n", nbits, nerrs, (float)nerrs/nbits);
+        fprintf(stderr, "ber_st: %d ber_en: %d nbits: %d nerr: %d BER: %4.3f\n", ber_st, ber_en,
+                nbits, nerrs, (float)nerrs/nbits);
     return 0;
 }

src/lpcnet_enc.c

@@ -56,7 +56,7 @@ int main(int argc, char **argv) {
     int   *m = pred_m;
     float *vq = pred_vq;
     int   logmag = 0;
-    int   direct_split = 0;
+    int   vq_type = LPCNET_PRED;
 
     fin = stdin;
     fout = stdout;
@@ -70,15 +70,16 @@ int main(int argc, char **argv) {
         {"numstages",    required_argument, 0, 'n'},
         {"pitchquant",   required_argument, 0, 'o'},
         {"pred",         required_argument, 0, 'p'},
-        {"directsplit",  no_argument,       0, 's'},
+        {"split",        no_argument,       0, 's'},
+        {"indexopt",     no_argument,       0, 'x'},
         {"verbose",      no_argument,       0, 'v'},
         {0, 0, 0, 0}
     };
 
     int   c;
     int opt_index = 0;
 
-    while ((c = getopt_long (argc, argv, "d:n:o:p:svi:u:", long_options, &opt_index)) != -1) {
+    while ((c = getopt_long (argc, argv, "d:n:o:p:sxvi:u:", long_options, &opt_index)) != -1) {
         switch (c) {
 	case 'i':
             if ((fin = fopen(optarg, "rb")) == NULL) {
@@ -109,9 +110,14 @@ int main(int argc, char **argv) {
             fprintf(stderr, "pred = %f\n", pred);
             break;
         case 's':
-            direct_split = 1;
+            vq_type = LPCNET_DIRECT_SPLIT;
             m = direct_split_m; vq = direct_split_vq; pred = 0.0; logmag = 1; weight = 1.0;
-            fprintf(stderr, "split VQ\n");
+            fprintf(stderr, "direct split VQ\n");
+            break;
+        case 'x':
+            vq_type = LPCNET_DIRECT_SPLIT_INDEX_OPT;
+            m = direct_split_indopt_m; vq = direct_split_indopt_vq; pred = 0.0; logmag = 1; weight = 1.0;
+            fprintf(stderr, "index optimised direct split VQ\n");
             break;
         case 'v':
             lpcnet_verbose = 1;
@@ -120,13 +126,15 @@ int main(int argc, char **argv) {
             fprintf(stderr,"usage: %s [Options]:\n  [-d --decimation 1/2/3...]\n", argv[0]);
             fprintf(stderr,"  [-i --infile]\n  [-u --outfile]\n");
             fprintf(stderr,"  [-n --numstages]\n  [-o --pitchbits nBits]\n");
-            fprintf(stderr,"  [-p --pred predCoff] [-s --split]\n");
+            fprintf(stderr,"  [-p --pred predCoff]  Predictive quantiser prediction coeff\n");
+            fprintf(stderr,"  [-s --directsplit]    Use direct split quaniser\n");
+            fprintf(stderr,"  [-x --indexopt]       Use index optimised direct split quantiser\n");
             fprintf(stderr,"  [-v --verbose]\n");
             exit(1);
         }
     }
 
-    LPCNetFreeDV *lf = lpcnet_freedv_create(direct_split);
+    LPCNetFreeDV *lf = lpcnet_freedv_create(vq_type);
     LPCNET_QUANT *q = lf->q;
 
     q->weight = weight; q->pred = pred; q->mbest = mbest_survivors;

src/lpcnet_freedv.h

@@ -12,9 +12,14 @@
   extern "C" {
 #endif
 
+// possible vq_type values in lpcnet_freedv_create()
+#define LPCNET_PRED                   0
+#define LPCNET_DIRECT_SPLIT           1
+#define LPCNET_DIRECT_SPLIT_INDEX_OPT 2
+
 typedef struct LPCNetFreeDV LPCNetFreeDV;
 
-LPCNetFreeDV* lpcnet_freedv_create(int direct_split);
+LPCNetFreeDV* lpcnet_freedv_create(int vq_type);
 void lpcnet_freedv_destroy(LPCNetFreeDV *lf);
 void lpcnet_enc(LPCNetFreeDV *lf, short *pcm, char *frame);
 void lpcnet_dec(LPCNetFreeDV *lf, char *frame, short* pcm);

src/lpcnet_quant.c

@@ -30,13 +30,24 @@ int lpcnet_verbose = 0;
 
 static int quantise(const float * cb, float vec[], float w[], int k, int m, float *se);
 
-LPCNET_QUANT *lpcnet_quant_create(int direct_split) {
+LPCNET_QUANT *lpcnet_quant_create(int vq_type) {
+    assert((vq_type == LPCNET_PRED) || (vq_type == LPCNET_DIRECT_SPLIT) ||
+           (vq_type == LPCNET_DIRECT_SPLIT_INDEX_OPT));
     LPCNET_QUANT *q = (LPCNET_QUANT*)malloc(sizeof(LPCNET_QUANT));
     if (q == NULL) return NULL;
-    if (direct_split) {
+    if ((vq_type == LPCNET_DIRECT_SPLIT) || (vq_type == LPCNET_DIRECT_SPLIT_INDEX_OPT)) {
         q->weight = 1.0; q->pred = 0.0; 
         q->mbest = DEFAULT_MBEST; q->pitch_bits = DEFAULT_PITCH_BITS; q->dec = DEFAULT_DEC;
-        q->num_stages = direct_split_num_stages; q->vq = direct_split_vq; q->m = direct_split_m; q->logmag = 1;
+        q->logmag = 1;
+        if (vq_type == LPCNET_DIRECT_SPLIT) {
+            q->num_stages = direct_split_num_stages;
+            q->m = direct_split_m; 
+            q->vq = direct_split_vq;
+        } else {
+            q->num_stages = direct_split_indopt_num_stages;
+            q->m = direct_split_indopt_m; 
+            q->vq = direct_split_indopt_vq;
+        }
     }
     else {
         q->weight = DEFAULT_WEIGHT; q->pred = DEFAULT_PRED; 
@@ -125,7 +136,8 @@ void quant_pred_mbest(float vec_out[],
                       int num_stages,
                       float vq[],
                       int m[], int k,
-                      int mbest_survivors)
+                      int mbest_survivors,
+                      float ber)
 {
     float err[k], w[k], se1;
     int i,j,s,s1,ind;
@@ -194,6 +206,16 @@ void quant_pred_mbest(float vec_out[],
     if (lpcnet_fsv != NULL) fprintf(lpcnet_fsv, "%f\t%f\t", vec_in[0],sqrt(se1));
     if (lpcnet_verbose) fprintf(stderr, "    se1: %f\n", se1);
 
+    if (ber > 0.0) {
+        /* optionally insert random errors in indexes to test index optimisation */
+        for (s=0; s<num_stages; s++) {
+            int log2m = roundf(log2(m[s]));
+            for(int b=0; b<log2m; b++)
+                if ((float)rand()/RAND_MAX < ber)
+                    indexes[s] ^= 1<<b;
+        }
+    }
+
     quant_pred_output(vec_out, indexes, err, pred, num_stages, vq, k);
 
     for(i=0; i<num_stages; i++)
@@ -365,7 +387,7 @@ int lpcnet_features_to_frame(LPCNET_QUANT *q, float features[], char frame[]) {
 
     /* non-interpolated frame ----------------------------------------*/
 
-    quant_pred_mbest(q->features_quant, indexes, features, q->pred, q->num_stages, q->vq, q->m, k, q->mbest);
+    quant_pred_mbest(q->features_quant, indexes, features, q->pred, q->num_stages, q->vq, q->m, k, q->mbest, 0.0);
     pitch_ind = pitch_encode(features[2*NB_BANDS], q->pitch_bits);
     pitch_gain_ind =  pitch_gain_encode(features[2*NB_BANDS+1]);
     pack_frame(q->num_stages, q->m, indexes, q->pitch_bits, pitch_ind, pitch_gain_ind, frame);