Merge branch 'short-clip'

NEWS

+Overview of Changes in audiowmark-0.3.0:
+
+* replace padding at start with a partial B block
+* add algorithm for decoding the watermark from short clips
+
 Overview of Changes in audiowmark-0.2.1:
 
 * add limiter to avoid clipping during watermark generation

configure.ac

-AC_INIT([audiowmark], [0.2.1])
+AC_INIT([audiowmark], [0.3.0])
 AC_CONFIG_SRCDIR([src/audiowmark.cc])
 AC_CONFIG_AUX_DIR([build-aux])
 AC_CONFIG_MACRO_DIR([m4])

src/.gitignore

@@ -7,3 +7,4 @@ testconvcode
 testmp3
 testrandom
 teststream
+testlimiter

src/audiowmark.cc

@@ -441,6 +441,45 @@ test_snr (const string& orig_file, const string& wm_file)
   return 0;
 }
 
+int
+test_clip (const string& in_file, const string& out_file, int seed, int time_seconds)
+{
+  WavData in_data;
+  Error err = in_data.load (in_file);
+  if (err)
+    {
+      error ("audiowmark: error loading %s: %s\n", in_file.c_str(), err.message());
+      return 1;
+    }
+  bool done = false;
+  Random rng (seed, /* there is no stream for this test */ Random::Stream::data_up_down);
+  size_t start_point, end_point;
+  do
+    {
+      // this is unbiased only if 2 * block_size + time_seconds is smaller than overall file length
+      const size_t values_per_block = (mark_sync_frame_count() + mark_data_frame_count()) * Params::frame_size * in_data.n_channels();
+      start_point = 2 * values_per_block * (double(rng()) / UINT64_MAX);
+      start_point /= in_data.n_channels();
+
+      end_point = start_point + time_seconds * in_data.sample_rate();
+      if (end_point < in_data.n_values() / in_data.n_channels())
+        done = true;
+    }
+  while (!done);
+  //printf ("%.3f %.3f\n", start_point / double (in_data.sample_rate()), end_point / double (in_data.sample_rate()));
+
+  vector<float> out_signal (in_data.samples().begin() + start_point * in_data.n_channels(),
+                            in_data.samples().begin() + end_point * in_data.n_channels());
+  WavData out_wav_data (out_signal, in_data.n_channels(), in_data.sample_rate(), in_data.bit_depth());
+  err = out_wav_data.save (out_file);
+  if (err)
+    {
+      error ("audiowmark: error saving %s: %s\n", out_file.c_str(), err.message());
+      return 1;
+    }
+  return 0;
+}
+
 int
 gen_key (const string& outfile)
 {
@@ -495,6 +534,10 @@ main (int argc, char **argv)
     {
       test_snr (argv[2], argv[3]);
     }
+  else if (op == "test-clip" && argc == 6)
+    {
+      test_clip (argv[2], argv[3], atoi (argv[4]), atoi (argv[5]));
+    }
   else if (op == "gen-key" && argc == 3)
     {
       return gen_key (argv[2]);

src/ber-test.sh

@@ -16,6 +16,9 @@ fi
 if [ "x$AWM_FILE" == "x" ]; then
   AWM_FILE=t
 fi
+if [ "x$AWM_MULTI_CLIP" == "x" ]; then
+  AWM_MULTI_CLIP=1
+fi
 
 {
   if [ "x$AWM_SET" == "xsmall" ]; then
@@ -47,8 +50,8 @@ do
       PATTERN=4e1243bd22c66e76c2ba9eddc1f91394
     fi
     echo in_pattern $PATTERN
-    echo in_flags $AWM_PARAMS --test-key $SEED
-    audiowmark add "$i" ${AWM_FILE}.wav $PATTERN $AWM_PARAMS --test-key $SEED >/dev/null
+    echo in_flags $AWM_PARAMS $AWM_PARAMS_ADD --test-key $SEED
+    audiowmark add "$i" ${AWM_FILE}.wav $PATTERN $AWM_PARAMS $AWM_PARAMS_ADD --test-key $SEED --quiet
     if [ "x$AWM_RAND_CUT" != x ]; then
       CUT=$RANDOM
       audiowmark cut-start "${AWM_FILE}.wav" "${AWM_FILE}.wav" $CUT
@@ -91,7 +94,15 @@ do
       exit 1
     fi
     echo
-    if [ "x$AWM_REPORT" == "xtruncv" ]; then
+    if [ "x${AWM_CLIP}" != "x" ]; then
+      for CLIP in $(seq $AWM_MULTI_CLIP)
+      do
+        audiowmark test-clip $OUT_FILE ${OUT_FILE}.clip.wav $((CLIP_SEED++)) $AWM_CLIP --test-key $SEED
+        audiowmark cmp ${OUT_FILE}.clip.wav $PATTERN $AWM_PARAMS --test-key $SEED $TEST_CUT_ARGS
+        rm ${OUT_FILE}.clip.wav
+        echo
+      done
+    elif [ "x$AWM_REPORT" == "xtruncv" ]; then
       for TRUNC in $AWM_TRUNCATE
       do
         audiowmark cmp $OUT_FILE $PATTERN $AWM_PARAMS --test-key $SEED $TEST_CUT_ARGS --test-truncate $TRUNC | sed "s/^/$TRUNC /g"

src/gen-short-clip-adoc.sh

+#!/bin/bash
+
+STRENGTHS="10 15 20 30"
+STR_CLIPS="5 10 15 20 25 30"
+MAIN_CLIPS="5 10 15 20 30 40 50 60"
+
+echo ".performance-by-clip-length"
+echo '[frame="topbot",options="header",cols="<2,8*>1"]'
+echo '|=========================='
+echo -n "| Quality "
+for CLIP in $MAIN_CLIPS
+do
+  echo -n "| $CLIP"
+done
+echo
+
+for TEST in mp3-256 mp3-128 double-mp3-128 ogg-128
+do
+  echo -n "| $TEST "
+  for CLIP in $MAIN_CLIPS
+  do
+    cat main-$TEST-*-$CLIP | awk '{bad += $1; n += $2} END {if (n==0) n=1;fer=100.0*bad/n; bold=fer>0?"*":" ";printf ("| %s%.2f%s", bold, fer, bold)}'
+  done
+  echo
+done
+echo
+echo '|=========================='
+
+echo ".effects-of-watermarking-strength"
+echo '[frame="topbot",options="header",cols="<1,7*>1"]'
+echo '|=========================='
+echo -n "| Strength | SNR "
+for CLIP in $STR_CLIPS
+do
+  echo -n "| $CLIP"
+done
+echo
+
+for STRENGTH in $STRENGTHS
+do
+  echo -n "| $STRENGTH "
+  cat snr-$STRENGTH | awk '{printf ("| %.2f ", $1);}'
+  for CLIP in $STR_CLIPS
+  do
+    cat str-$STRENGTH-mp3-*-$CLIP | awk '{bad += $1; n += $2} END {if (n==0) n=1;fer=100.0*bad/n; bold=fer>0?"*":" ";printf ("| %s%.2f%s", bold, fer, bold)}'
+  done
+  echo
+done
+echo
+echo '|=========================='

src/gen-short-clip-mk.sh

+#!/bin/bash
+
+SEEDS=$(seq 5)
+STRENGTHS="10 15 20 30"
+STR_CLIPS="5 10 15 20 25 30"
+MAIN_CLIPS="5 10 15 20 30 40 50 60"
+MULTI_CLIP=4
+
+echo -n "all:"
+
+for STRENGTH in $STRENGTHS
+do
+  FILE="snr-$STRENGTH"
+  echo -n " $FILE"
+done
+
+for SEED in $SEEDS
+do
+  for STRENGTH in $STRENGTHS
+  do
+    for CLIP in $STR_CLIPS
+    do
+      FILE="str-$STRENGTH-mp3-$SEED-$CLIP"
+      echo -n " $FILE"
+    done
+  done
+
+  for CLIP in $MAIN_CLIPS
+  do
+    echo -n " main-mp3-256-$SEED-$CLIP main-mp3-128-$SEED-$CLIP main-ogg-128-$SEED-$CLIP main-double-mp3-128-$SEED-$CLIP"
+  done
+done
+
+echo
+echo
+
+for SEED in $SEEDS
+do
+  for STRENGTH in $STRENGTHS
+  do
+    for CLIP in $STR_CLIPS
+    do
+      FILE="str-$STRENGTH-mp3-$SEED-$CLIP"
+      echo "$FILE:"
+      echo -e "\t( cd ..; AWM_RAND_PATTERN=1 AWM_SET=huge2 AWM_PARAMS_ADD='--strength $STRENGTH' AWM_CLIP='$CLIP' AWM_MULTI_CLIP='$MULTI_CLIP' AWM_SEEDS=$SEED AWM_FILE='t-$FILE' ber-test.sh mp3 128 ) >x$FILE"
+      echo -e "\tmv x$FILE $FILE"
+      echo
+    done
+  done
+done
+
+for STRENGTH in $STRENGTHS
+do
+  FILE="snr-$STRENGTH"
+  echo "$FILE:"
+  echo -e "\t( cd ..; AWM_SET=huge2 AWM_PARAMS='--strength $STRENGTH' snr.sh ) >x$FILE"
+  echo -e "\tmv x$FILE $FILE"
+  echo
+done
+
+for SEED in $SEEDS
+do
+  for CLIP in $MAIN_CLIPS
+  do
+    FILE="main-mp3-256-$SEED-$CLIP"
+    echo "$FILE:"
+    echo -e "\t( cd ..; AWM_RAND_PATTERN=1 AWM_SET=huge2 AWM_CLIP='$CLIP' AWM_MULTI_CLIP='$MULTI_CLIP' AWM_SEEDS=$SEED AWM_FILE='t-$FILE' ber-test.sh mp3 256 ) >x$FILE"
+    echo -e "\tmv x$FILE $FILE"
+    echo
+
+    FILE="main-mp3-128-$SEED-$CLIP"
+    echo "$FILE:"
+    echo -e "\t( cd ..; AWM_RAND_PATTERN=1 AWM_SET=huge2 AWM_CLIP='$CLIP' AWM_MULTI_CLIP='$MULTI_CLIP' AWM_SEEDS=$SEED AWM_FILE='t-$FILE' ber-test.sh mp3 128 ) >x$FILE"
+    echo -e "\tmv x$FILE $FILE"
+    echo
+
+    FILE="main-ogg-128-$SEED-$CLIP"
+    echo "$FILE:"
+    echo -e "\t( cd ..; AWM_RAND_PATTERN=1 AWM_SET=huge2 AWM_CLIP='$CLIP' AWM_MULTI_CLIP='$MULTI_CLIP' AWM_SEEDS=$SEED AWM_FILE='t-$FILE' ber-test.sh ogg 128 ) >x$FILE"
+    echo -e "\tmv x$FILE $FILE"
+    echo
+
+    FILE="main-double-mp3-128-$SEED-$CLIP"
+    echo "$FILE:"
+    echo -e "\t( cd ..; AWM_RAND_PATTERN=1 AWM_SET=huge2 AWM_CLIP='$CLIP' AWM_MULTI_CLIP='$MULTI_CLIP' AWM_SEEDS=$SEED AWM_FILE='t-$FILE' ber-test.sh double-mp3 128 ) >x$FILE"
+    echo -e "\tmv x$FILE $FILE"
+    echo
+  done
+done

src/random.hh

@@ -13,7 +13,7 @@ public:
   enum class Stream {
     data_up_down = 1,
     sync_up_down = 2,
-    pad_up_down = 3,
+    pad_up_down = 3,   /* unused */
     mix = 4,
     bit_order = 5,
     frame_position = 6

src/short.mk

+all: short-60 short-60-mp3 \
+  short-50 short-50-mp3 \
+  short-40 short-40-mp3 \
+  short-30 short-30-mp3 \
+  short-20 short-20-mp3 \
+  short-10 short-10-mp3
+
+short-60:
+	AWM_FILE=t-short-60 AWM_CLIP=60 fer-test.sh 10 "" > tmp-short-60
+	mv tmp-short-60 short-60
+
+short-60-mp3:
+	AWM_FILE=t-short-60-mp3 AWM_CLIP=60 fer-test.sh 10 "" mp3 128 > tmp-short-60-mp3
+	mv tmp-short-60-mp3 short-60-mp3
+
+short-50:
+	AWM_FILE=t-short-50 AWM_CLIP=50 fer-test.sh 10 "" > tmp-short-50
+	mv tmp-short-50 short-50
+
+short-50-mp3:
+	AWM_FILE=t-short-50-mp3 AWM_CLIP=50 fer-test.sh 10 "" mp3 128 > tmp-short-50-mp3
+	mv tmp-short-50-mp3 short-50-mp3
+
+short-40:
+	AWM_FILE=t-short-40 AWM_CLIP=40 fer-test.sh 10 "" > tmp-short-40
+	mv tmp-short-40 short-40
+
+short-40-mp3:
+	AWM_FILE=t-short-40-mp3 AWM_CLIP=40 fer-test.sh 10 "" mp3 128 > tmp-short-40-mp3
+	mv tmp-short-40-mp3 short-40-mp3
+
+short-30:
+	AWM_FILE=t-short-30 AWM_CLIP=30 fer-test.sh 10 "" > tmp-short-30
+	mv tmp-short-30 short-30
+
+short-30-mp3:
+	AWM_FILE=t-short-30-mp3 AWM_CLIP=30 fer-test.sh 10 "" mp3 128 > tmp-short-30-mp3
+	mv tmp-short-30-mp3 short-30-mp3
+
+short-20:
+	AWM_FILE=t-short-20 AWM_CLIP=20 fer-test.sh 10 "" > tmp-short-20
+	mv tmp-short-20 short-20
+
+short-20-mp3:
+	AWM_FILE=t-short-20-mp3 AWM_CLIP=20 fer-test.sh 10 "" mp3 128 > tmp-short-20-mp3
+	mv tmp-short-20-mp3 short-20-mp3
+
+short-10:
+	AWM_FILE=t-short-10 AWM_CLIP=10 fer-test.sh 10 "" > tmp-short-10
+	mv tmp-short-10 short-10
+
+short-10-mp3:
+	AWM_FILE=t-short-10-mp3 AWM_CLIP=10 fer-test.sh 10 "" mp3 128 > tmp-short-10-mp3
+	mv tmp-short-10-mp3 short-10-mp3

src/snr.sh

@@ -4,5 +4,5 @@ PATTERN=4e1243bd22c66e76c2ba9eddc1f91394
 
 for i in test/T*
 do
-  echo $i $(audiowmark add $i t.wav $PATTERN $AWM_PARAMS --snr | grep SNR)
+  echo $i $(audiowmark add $i t.wav $PATTERN $AWM_PARAMS --snr 2>&1 | grep SNR)
 done | awk '{s += $3; n++} END { print s/n; }'

src/wmadd.cc

@@ -17,6 +17,7 @@
 using std::string;
 using std::vector;
 using std::complex;
+using std::max;
 
 enum class FrameMod : uint8_t {
   KEEP = 0,
@@ -121,20 +122,6 @@ mark_sync (vector<vector<FrameMod>>& frame_mod, int ab)
     }
 }
 
-static void
-init_pad_mod_vec (vector<vector<FrameMod>>& pad_mod_vec)
-{
-  UpDownGen up_down_gen (Random::Stream::pad_up_down);
-
-  for (size_t f = 0; f < Params::frames_pad_start; f++)
-    {
-      vector<FrameMod> mod (Params::max_band + 1);
-
-      prepare_frame_mod (up_down_gen, f, mod, 0);
-      pad_mod_vec.push_back (mod);
-    }
-}
-
 static void
 init_frame_mod_vec (vector<vector<FrameMod>>& frame_mod_vec, int ab, const vector<int>& bitvec)
 {
@@ -245,11 +232,8 @@ public:
  */
 class WatermarkGen
 {
-  enum State { PAD, WATERMARK } state = State::PAD;
-
   const int                 n_channels = 0;
-  int                       frame_number = 0;
-  int                       frame_bound = Params::frames_pad_start;
+  size_t                    frame_number = 0;
   int                       ab = 0;
   int                       m_data_blocks = 0;
 
@@ -257,7 +241,6 @@ class WatermarkGen
   WatermarkSynth            wm_synth;
 
   vector<int>               bitvec;
-  vector<vector<FrameMod>>  pad_mod_vec;
   vector<vector<FrameMod>>  frame_mod_vec_a;
   vector<vector<FrameMod>>  frame_mod_vec_b;
 public:
@@ -267,7 +250,12 @@ public:
     wm_synth (n_channels),
     bitvec (bitvec)
   {
-    init_pad_mod_vec (pad_mod_vec);
+    /* start writing a partial B-block as padding */
+    init_frame_mod_vec (frame_mod_vec_b, 1, bitvec);
+
+    assert (frame_mod_vec_b.size() > Params::frames_pad_start);
+    frame_number = frame_mod_vec_b.size() - Params::frames_pad_start;
+    ab = 1;
   }
   vector<float>
   run (const vector<float>& samples)
@@ -280,50 +268,29 @@ public:
     for (int ch = 0; ch < n_channels; ch++)
       fft_delta_spect.push_back (vector<complex<float>> (fft_out.back().size()));
 
-    if (state == State::PAD)
-      {
-        for (int ch = 0; ch < n_channels; ch++)
-          apply_frame_mod (pad_mod_vec[frame_number], fft_out[ch], fft_delta_spect[ch]);
-      }
-    else if (state == State::WATERMARK)
-      {
-        for (int ch = 0; ch < n_channels; ch++)
-          apply_frame_mod (ab ? frame_mod_vec_b[frame_number] : frame_mod_vec_a[frame_number], fft_out[ch], fft_delta_spect[ch]);
-      }
+    const vector<vector<FrameMod>>& frame_mod_vec = ab ? frame_mod_vec_b : frame_mod_vec_a;
+    for (int ch = 0; ch < n_channels; ch++)
+      apply_frame_mod (frame_mod_vec[frame_number], fft_out[ch], fft_delta_spect[ch]);
 
     frame_number++;
-    if (frame_number == frame_bound)
+    if (frame_number == frame_mod_vec.size())
       {
         frame_number = 0;
 
-        if (state == PAD)
-          {
-            state = WATERMARK;
-            frame_bound = mark_sync_frame_count() + mark_data_frame_count();
+        ab = (ab + 1) & 1; // write A|B|A|B|...
+        m_data_blocks++;
 
-            // initialize a-block frame mod vector here to minimize startup latency
-            assert (frame_mod_vec_a.empty());
-            init_frame_mod_vec (frame_mod_vec_a, 0, bitvec);
-          }
-        else if (state == WATERMARK)
-          {
-            ab = (ab + 1) & 1; // write A|B|A|B|...
-            frame_bound = mark_sync_frame_count() + mark_data_frame_count();
-            m_data_blocks++;
-
-            if (frame_mod_vec_b.empty())
-              {
-                // initialize b-block frame mod vector here to minimize startup latency
-                init_frame_mod_vec (frame_mod_vec_b, 1, bitvec);
-              }
-          }
+        // initialize A-block frame mod vector here to minimize startup latency
+        if (frame_mod_vec_a.empty())
+          init_frame_mod_vec (frame_mod_vec_a, 0, bitvec);
       }
     return wm_synth.run (fft_delta_spect);
   }
   int
   data_blocks() const
   {
-    return m_data_blocks;
+    // first block is padding - a partial B block
+    return max (m_data_blocks - 1, 0);
   }
 };