Update pitch_mark and pitch2sin

CMakeLists.txt

@@ -143,8 +143,8 @@ set(CC_SOURCES
     ${SOURCE_DIR}/generation/m_sequence_generation.cc
     ${SOURCE_DIR}/generation/nonrecursive_maximum_likelihood_parameter_generation.cc
     ${SOURCE_DIR}/generation/normal_distributed_random_value_generation.cc
+    ${SOURCE_DIR}/generation/periodic_waveform_generation.cc
     ${SOURCE_DIR}/generation/recursive_maximum_likelihood_parameter_generation.cc
-    ${SOURCE_DIR}/generation/sinusoidal_generation.cc
     ${SOURCE_DIR}/input/input_source_delay.cc
     ${SOURCE_DIR}/input/input_source_filling_magic_number.cc
     ${SOURCE_DIR}/input/input_source_from_array.cc

include/SPTK/generation/sinusoidal_generation.h → include/SPTK/generation/periodic_waveform_generation.h

@@ -14,31 +14,33 @@
 // limitations under the License.                                           //
 // ------------------------------------------------------------------------ //
 
-#ifndef SPTK_GENERATION_SINUSOIDAL_GENERATION_H_
-#define SPTK_GENERATION_SINUSOIDAL_GENERATION_H_
+#ifndef SPTK_GENERATION_PERIODIC_WAVEFORM_GENERATION_H_
+#define SPTK_GENERATION_PERIODIC_WAVEFORM_GENERATION_H_
 
 #include "SPTK/input/input_source_interpolation_with_magic_number.h"
 #include "SPTK/utils/sptk_utils.h"
 
 namespace sptk {
 
 /**
- * Generate sinusoidal sequence.
+ * Generate periodic waveform from pitch value.
  *
  * The input is a sequence of pitch value which can be either a continuous value
- * or a magic number. The output is the sinusoidal signal given the input
+ * or a magic number. The output is the periodic waveform generated by the input
  * sequence.
  */
-class SinusoidalGeneration {
+class PeriodicWaveformGeneration {
  public:
   /**
+   * @param[in] unvoiced_value Value on unvoiced region.
    * @param[in] strict If true, strictly drop sinusoidal in an unvoiced region.
    * @param[in] input_source Input source.
    */
-  explicit SinusoidalGeneration(
-      bool strict, InputSourceInterpolationWithMagicNumber* input_source);
+  PeriodicWaveformGeneration(
+      double unvoiced_value, bool strict,
+      InputSourceInterpolationWithMagicNumber* input_source);
 
-  virtual ~SinusoidalGeneration() {
+  virtual ~PeriodicWaveformGeneration() {
   }
 
   /**
@@ -49,16 +51,18 @@ class SinusoidalGeneration {
   }
 
   /**
-   * Get sinusoidal signal.
+   * Get periodic waveform.
    *
    * @param[out] sin Sine waveform (optional).
    * @param[out] cos Cosine waveform (optional).
+   * @param[out] sawtooth Sawtooth waveform (optional).
    * @param[out] pitch Pitch (optional).
    * @return True on success, false on failure.
    */
-  bool Get(double* sin, double* cos, double* pitch);
+  bool Get(double* sin, double* cos, double* sawtooth, double* pitch);
 
  private:
+  const double unvoiced_value_;
   const bool strict_;
 
   InputSourceInterpolationWithMagicNumber* input_source_;
@@ -72,9 +76,9 @@ class SinusoidalGeneration {
   double voiced_pitch_;
   bool extending_;
 
-  DISALLOW_COPY_AND_ASSIGN(SinusoidalGeneration);
+  DISALLOW_COPY_AND_ASSIGN(PeriodicWaveformGeneration);
 };
 
 }  // namespace sptk
 
-#endif  // SPTK_GENERATION_SINUSOIDAL_GENERATION_H_
+#endif  // SPTK_GENERATION_PERIODIC_WAVEFORM_GENERATION_H_

src/generation/sinusoidal_generation.cc → src/generation/periodic_waveform_generation.cc

@@ -14,16 +14,18 @@
 // limitations under the License.                                           //
 // ------------------------------------------------------------------------ //
 
-#include "SPTK/generation/sinusoidal_generation.h"
+#include "SPTK/generation/periodic_waveform_generation.h"
 
 #include <cmath>   // std::cos, std::sin
 #include <vector>  // std::vector
 
 namespace sptk {
 
-SinusoidalGeneration::SinusoidalGeneration(
-    bool strict, InputSourceInterpolationWithMagicNumber* input_source)
-    : strict_(strict),
+PeriodicWaveformGeneration::PeriodicWaveformGeneration(
+    double unvoiced_value, bool strict,
+    InputSourceInterpolationWithMagicNumber* input_source)
+    : unvoiced_value_(unvoiced_value),
+      strict_(strict),
       input_source_(input_source),
       is_valid_(true),
       phase_(0.0),
@@ -35,7 +37,8 @@ SinusoidalGeneration::SinusoidalGeneration(
   }
 }
 
-bool SinusoidalGeneration::Get(double* sin, double* cos, double* pitch) {
+bool PeriodicWaveformGeneration::Get(double* sin, double* cos, double* sawtooth,
+                                     double* pitch) {
   if (!is_valid_) {
     return false;
   }
@@ -64,10 +67,13 @@ bool SinusoidalGeneration::Get(double* sin, double* cos, double* pitch) {
   if (input_source_->GetMagicNumber() == pitch_in_current_point) {
     phase_ = 0.0;
     if (sin) {
-      *sin = 0.0;
+      *sin = unvoiced_value_;
     }
     if (cos) {
-      *cos = 0.0;
+      *cos = unvoiced_value_;
+    }
+    if (sawtooth) {
+      *sawtooth = unvoiced_value_;
     }
     return true;
   }
@@ -78,6 +84,9 @@ bool SinusoidalGeneration::Get(double* sin, double* cos, double* pitch) {
   if (cos) {
     *cos = std::cos(phase_);
   }
+  if (sawtooth) {
+    *sawtooth = phase_ / sptk::kPi - 1.0;
+  }
 
   if (!strict_) {
     extending_ = true;

src/main/pitch2sin.cc

@@ -20,19 +20,20 @@
 #include <sstream>   // std::ostringstream
 
 #include "GETOPT/ya_getopt.h"
-#include "SPTK/generation/sinusoidal_generation.h"
+#include "SPTK/generation/periodic_waveform_generation.h"
 #include "SPTK/input/input_source_from_stream.h"
 #include "SPTK/input/input_source_interpolation_with_magic_number.h"
 #include "SPTK/utils/sptk_utils.h"
 
 namespace {
 
-enum OutputFormats { kSine = 0, kCosine, kNumOutputFormats };
+enum OutputFormats { kSine = 0, kCosine, kSawtooth, kNumOutputFormats };
 
 const int kDefaultFramePeriod(100);
 const int kDefaultInterpolationPeriod(1);
 const OutputFormats kDefaultOutputFormat(kSine);
 const bool kDefaultStrictFlag(false);
+const double kDefaultUnvoicedValue(0.0);
 const double kMagicNumberForUnvoicedFrame(0.0);
 
 void PrintUsage(std::ostream* stream) {
@@ -45,9 +46,12 @@ void PrintUsage(std::ostream* stream) {
   *stream << "  options:" << std::endl;
   *stream << "       -p p  : frame period         (   int)[" << std::setw(5) << std::right << kDefaultFramePeriod         << "][ 1 <= p <=     ]" << std::endl;  // NOLINT
   *stream << "       -i i  : interpolation period (   int)[" << std::setw(5) << std::right << kDefaultInterpolationPeriod << "][ 0 <= i <= p/2 ]" << std::endl;  // NOLINT
-  *stream << "       -o o  : output format        (   int)[" << std::setw(5) << std::right << kDefaultOutputFormat        << "][ 0 <= o <= 1   ]" << std::endl;  // NOLINT
+  *stream << "       -o o  : output format        (   int)[" << std::setw(5) << std::right << kDefaultOutputFormat        << "][ 0 <= o <= 2   ]" << std::endl;  // NOLINT
   *stream << "                 0 (sine)" << std::endl;
   *stream << "                 1 (cosine)" << std::endl;
+  *stream << "                 2 (sawtooth)" << std::endl;
+  *stream << "       -u u  : value on unvoiced    (double)[" << std::setw(5) << std::right << kDefaultUnvoicedValue       << "][   <= u <=     ]" << std::endl;  // NOLINT
+  *stream << "               region" << std::endl;
   *stream << "       -s    : strictly drop signal (  bool)[" << std::setw(5) << std::right << sptk::ConvertBooleanToString(kDefaultStrictFlag) << "]" << std::endl;  // NOLINT
   *stream << "               in unvoiced region" << std::endl;
   *stream << "       -h    : print this message" << std::endl;
@@ -77,6 +81,9 @@ void PrintUsage(std::ostream* stream) {
  *   - output format
  *     \arg @c 0 sine
  *     \arg @c 1 cosine
+ *     \arg @c 2 sawtooth
+ * - @b -u @e double
+ *   - value on unvoiced region
  * - @b -s
  *   - strictly drop sinusoidal in unvoiced region
  * - @b infile @e str
@@ -98,10 +105,11 @@ int main(int argc, char* argv[]) {
   int frame_period(kDefaultFramePeriod);
   int interpolation_period(kDefaultInterpolationPeriod);
   OutputFormats output_format(kDefaultOutputFormat);
+  double unvoiced_value(kDefaultUnvoicedValue);
   bool strict(kDefaultStrictFlag);
 
   for (;;) {
-    const int option_char(getopt_long(argc, argv, "p:i:o:sh", NULL, NULL));
+    const int option_char(getopt_long(argc, argv, "p:i:o:u:sh", NULL, NULL));
     if (-1 == option_char) break;
 
     switch (option_char) {
@@ -142,6 +150,15 @@ int main(int argc, char* argv[]) {
         output_format = static_cast<OutputFormats>(tmp);
         break;
       }
+      case 'u': {
+        if (!sptk::ConvertStringToDouble(optarg, &unvoiced_value)) {
+          std::ostringstream error_message;
+          error_message << "The argument for the -u option must be a number";
+          sptk::PrintErrorMessage("pitch2sin", error_message);
+          return 1;
+        }
+        break;
+      }
       case 's': {
         strict = true;
         break;
@@ -205,22 +222,23 @@ int main(int argc, char* argv[]) {
     return 1;
   }
 
-  sptk::SinusoidalGeneration sinusoidal_generation(
-      strict, &input_source_interpolation_with_magic_number);
-  if (!sinusoidal_generation.IsValid()) {
+  sptk::PeriodicWaveformGeneration waveform_generation(
+      unvoiced_value, strict, &input_source_interpolation_with_magic_number);
+  if (!waveform_generation.IsValid()) {
     std::ostringstream error_message;
-    error_message << "Failed to initialize SinusoidalGeneration";
+    error_message << "Failed to initialize PeriodicWaveformGeneration";
     sptk::PrintErrorMessage("pitch2sin", error_message);
     return 1;
   }
 
-  double sinusoidal;
-  while (sinusoidal_generation.Get(
-      kSine == output_format ? &sinusoidal : NULL,
-      kCosine == output_format ? &sinusoidal : NULL, NULL)) {
-    if (!sptk::WriteStream(sinusoidal, &std::cout)) {
+  double signal;
+  while (waveform_generation.Get(kSine == output_format ? &signal : NULL,
+                                 kCosine == output_format ? &signal : NULL,
+                                 kSawtooth == output_format ? &signal : NULL,
+                                 NULL)) {
+    if (!sptk::WriteStream(signal, &std::cout)) {
       std::ostringstream error_message;
-      error_message << "Failed to write sinusoidal";
+      error_message << "Failed to write a periodic waveform";
       sptk::PrintErrorMessage("pitch2sin", error_message);
       return 1;
     }

src/main/pitch_mark.cc

@@ -35,8 +35,7 @@ enum OutputFormats {
   kPositionInSamples,
   kSine,
   kCosine,
-  kBipolarSawtooth,
-  kUnipolarSawtooth,
+  kSawtooth,
   kNumOutputFormats
 };
 
@@ -45,6 +44,7 @@ const double kDefaultLowerF0(60.0);
 const double kDefaultUpperF0(240.0);
 const double kDefaultVoicingThreshold(0.9);
 const OutputFormats kDefaultOutputFormat(kBinarySequence);
+const double kDefaultUnvoicedValue(0.0);
 
 void PrintUsage(std::ostream* stream) {
   // clang-format off
@@ -60,14 +60,14 @@ void PrintUsage(std::ostream* stream) {
   *stream << "       -H H  : maximum fundamental frequency (double)[" << std::setw(5) << std::right << kDefaultUpperF0          << "][    L <  H <  500*s ]" << std::endl;  // NOLINT
   *stream << "               to search for [Hz]" << std::endl;
   *stream << "       -t t  : voicing threshold             (double)[" << std::setw(5) << std::right << kDefaultVoicingThreshold << "][ -0.5 <= t <= 1.6   ]" << std::endl;  // NOLINT
-  *stream << "       -o o  : output format                 (   int)[" << std::setw(5) << std::right << kDefaultOutputFormat     << "][    0 <= o <= 4     ]" << std::endl;  // NOLINT
+  *stream << "       -o o  : output format                 (   int)[" << std::setw(5) << std::right << kDefaultOutputFormat     << "][    0 <= o <= 5     ]" << std::endl;  // NOLINT
   *stream << "                 0 (binary sequence)" << std::endl;
   *stream << "                 1 (position in seconds)" << std::endl;
   *stream << "                 2 (position in samples)" << std::endl;
   *stream << "                 3 (sine waveform)" << std::endl;
   *stream << "                 4 (cosine waveform)" << std::endl;
-  *stream << "                 5 (sawtooth waveform (bipolar))" << std::endl;
-  *stream << "                 6 (sawtooth waveform (unipolar))" << std::endl;
+  *stream << "                 5 (sawtooth waveform)" << std::endl;
+  *stream << "       -u u  : value on unvoiced region      (double)[" << std::setw(5) << std::right << kDefaultUnvoicedValue    << "][      <= u <=       ]" << std::endl;  // NOLINT
   *stream << "       -h    : print this message" << std::endl;
   *stream << "  infile:" << std::endl;
   *stream << "       waveform                              (double)[stdin]" << std::endl;  // NOLINT
@@ -77,6 +77,7 @@ void PrintUsage(std::ostream* stream) {
   *stream << "       if t is raised, the number of pitch marks increase" << std::endl;  // NOLINT
   *stream << "       the value of t should be in the recommended range but values outside the range can be given" << std::endl;  // NOLINT
   *stream << "       if o = 0, value 1 or -1 indicating pitch mark is outputted considering polarity" << std::endl;  // NOLINT
+  *stream << "       -u option is valid only o >= 3" << std::endl;
   *stream << std::endl;
   *stream << " SPTK: version " << sptk::kVersion << std::endl;
   *stream << std::endl;
@@ -103,8 +104,9 @@ void PrintUsage(std::ostream* stream) {
  *     @arg @c 2 position in samples
  *     @arg @c 3 sine waveform
  *     @arg @c 4 cosine waveform
- *     @arg @c 5 bipolar sawtooth waveform
- *     @arg @c 6 unipolar sawtooth waveform
+ *     @arg @c 5 sawtooth waveform
+ * - @b -u @e double
+ *   - value on unvoiced region
  * - @b infile @e str
  *   - double-type waveform
  * - @b stdout
@@ -128,9 +130,10 @@ int main(int argc, char* argv[]) {
   double upper_f0(kDefaultUpperF0);
   double voicing_threshold(kDefaultVoicingThreshold);
   OutputFormats output_format(kDefaultOutputFormat);
+  double unvoiced_value(kDefaultUnvoicedValue);
 
   for (;;) {
-    const int option_char(getopt_long(argc, argv, "s:L:H:t:o:h", NULL, NULL));
+    const int option_char(getopt_long(argc, argv, "s:L:H:t:o:u:h", NULL, NULL));
     if (-1 == option_char) break;
 
     switch (option_char) {
@@ -193,6 +196,15 @@ int main(int argc, char* argv[]) {
         output_format = static_cast<OutputFormats>(tmp);
         break;
       }
+      case 'u': {
+        if (!sptk::ConvertStringToDouble(optarg, &unvoiced_value)) {
+          std::ostringstream error_message;
+          error_message << "The argument for the -u option must be a number";
+          sptk::PrintErrorMessage("pitch_mark", error_message);
+          return 1;
+        }
+        break;
+      }
       case 'h': {
         PrintUsage(&std::cout);
         return 0;
@@ -340,8 +352,7 @@ int main(int argc, char* argv[]) {
     }
     case kSine:
     case kCosine:
-    case kBipolarSawtooth:
-    case kUnipolarSawtooth: {
+    case kSawtooth: {
       for (int n(0), i(0); n <= num_pitch_marks; ++n) {
         const int next_pitch_mark(
             n < num_pitch_marks ? static_cast<int>(std::round(pitch_mark[n]))
@@ -354,7 +365,7 @@ int main(int argc, char* argv[]) {
           }
         }
 
-        // Output sinusoidal sequence.
+        // Output periodic sequence.
         if (i < j) {
           const double sum_f0(
               std::accumulate(f0.begin() + i, f0.begin() + j, 0.0));
@@ -372,21 +383,17 @@ int main(int argc, char* argv[]) {
                 value = std::cos(phase);
                 break;
               }
-              case kBipolarSawtooth: {
+              case kSawtooth: {
                 value = std::fmod(phase, sptk::kTwoPi) / sptk::kPi - 1.0;
                 break;
               }
-              case kUnipolarSawtooth: {
-                value = std::fmod(phase, sptk::kTwoPi) / sptk::kTwoPi;
-                break;
-              }
               default: {
                 return 1;
               }
             }
             if (!sptk::WriteStream(binary_polarity * value, &std::cout)) {
               std::ostringstream error_message;
-              error_message << "Failed to write sinusoidal sequence";
+              error_message << "Failed to write periodic sequence";
               sptk::PrintErrorMessage("pitch_mark", error_message);
               return 1;
             }
@@ -396,9 +403,9 @@ int main(int argc, char* argv[]) {
 
         // Output unvoiced sequence.
         for (int k(j); k < next_pitch_mark; ++k) {
-          if (!sptk::WriteStream(0.0, &std::cout)) {
+          if (!sptk::WriteStream(unvoiced_value, &std::cout)) {
             std::ostringstream error_message;
-            error_message << "Failed to write sinusoidal sequence";
+            error_message << "Failed to write periodic sequence";
             sptk::PrintErrorMessage("pitch_mark", error_message);
             return 1;
           }