Merge pull request #78 from sp-nitech/pitch_mark

Update pitch_mark

Author: takenori-y

include/SPTK/generation/excitation_generation.h

@@ -32,12 +32,23 @@ namespace sptk {
  */
 class ExcitationGeneration {
  public:
+  /**
+   * Normalization type.
+   */
+  enum NormalizationType {
+    kNone = 0,
+    kPower,
+    kMagnitude,
+    kNumNormalizationTypes
+  };
+
   /**
    * @param[in] input_source Input source.
    * @param[in] random_generation Random value generator.
    */
   ExcitationGeneration(InputSourceInterpolationWithMagicNumber* input_source,
-                       RandomGenerationInterface* random_generation);
+                       RandomGenerationInterface* random_generation,
+                       NormalizationType normalization_type = kPower);
 
   virtual ~ExcitationGeneration() {
   }
@@ -63,6 +74,7 @@ class ExcitationGeneration {
  private:
   InputSourceInterpolationWithMagicNumber* input_source_;
   RandomGenerationInterface* random_generation_;
+  const NormalizationType normalization_type_;
 
   bool is_valid_;
 

src/analysis/goertzel_analysis.cc

@@ -16,7 +16,7 @@
 
 #include "SPTK/analysis/goertzel_analysis.h"
 
-#include <cmath>    // std::round
+#include <cmath>    // std::cos, std::round, std::sin
 #include <cstddef>  // std::size_t
 #include <vector>   // std::vector
 

src/generation/excitation_generation.cc

@@ -23,9 +23,11 @@ namespace sptk {
 
 ExcitationGeneration::ExcitationGeneration(
     InputSourceInterpolationWithMagicNumber* input_source,
-    RandomGenerationInterface* random_generation)
+    RandomGenerationInterface* random_generation,
+    NormalizationType normalization_type)
     : input_source_(input_source),
       random_generation_(random_generation),
+      normalization_type_(normalization_type),
       is_valid_(true),
       phase_(1.0) {
   if (NULL == input_source_ || !input_source_->IsValid()) {
@@ -79,7 +81,15 @@ bool ExcitationGeneration::Get(double* excitation, double* pulse, double* noise,
   double pulse_in_current_point;
   if (1.0 <= phase_) {
     phase_ -= 1.0;
-    pulse_in_current_point = std::sqrt(pitch_in_current_point);
+    if (kNone == normalization_type_) {
+      pulse_in_current_point = 1.0;
+    } else if (kPower == normalization_type_) {
+      pulse_in_current_point = std::sqrt(pitch_in_current_point);
+    } else if (kMagnitude == normalization_type_) {
+      pulse_in_current_point = pitch_in_current_point;
+    } else {
+      return false;
+    }
   } else {
     pulse_in_current_point = 0.0;
   }

src/main/excite.cc

@@ -35,6 +35,8 @@ enum NoiseType { kZero = 0, kGaussian, kMSequence, kNumNoiseTypes };
 const int kDefaultFramePeriod(100);
 const int kDefaultInterpolationPeriod(1);
 const NoiseType kDefaultNoiseType(NoiseType::kMSequence);
+const sptk::ExcitationGeneration::NormalizationType kDefaultNormalizationType(
+    sptk::ExcitationGeneration::NormalizationType::kPower);
 const int kDefaultSeed(1);
 const double kMagicNumberForUnvoicedFrame(0.0);
 
@@ -48,6 +50,10 @@ 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 << "       -N N  : normalization type         (   int)[" << std::setw(5) << std::right << kDefaultNormalizationType   << "][ 0 <= N <= 2   ]" << std::endl;  // NOLINT
+  *stream << "                 0 (none)" << std::endl;
+  *stream << "                 1 (power)" << std::endl;
+  *stream << "                 2 (magnitude)" << std::endl;
   *stream << "       -n n  : noise type                 (   int)[" << std::setw(5) << std::right << kDefaultNoiseType           << "][ 0 <= n <= 2   ]" << std::endl;  // NOLINT
   *stream << "                 0 (none)" << std::endl;
   *stream << "                 1 (Gaussian)" << std::endl;
@@ -60,7 +66,7 @@ void PrintUsage(std::ostream* stream) {
   *stream << "       excitation                         (double)" << std::endl;
   *stream << "  notice:" << std::endl;
   *stream << "       if i = 0, don't interpolate pitch" << std::endl;
-  *stream << "       s is valid only if n = 1" << std::endl;
+  *stream << "       -s option is valid only if n = 1" << std::endl;
   *stream << "       magic number for unvoiced frame is " << kMagicNumberForUnvoicedFrame << std::endl;  // NOLINT
   *stream << std::endl;
   *stream << " SPTK: version " << sptk::kVersion << std::endl;
@@ -77,6 +83,11 @@ void PrintUsage(std::ostream* stream) {
  *   - frame_period @f$(1 \le P)@f$
  * - @b -i @e int
  *   - interpolation period @f$(0 \le I \le P/2)@f$
+ * - @b -N @e int
+ *   - normalization type
+ *     \arg @c 0 none
+ *     \arg @c 1 power
+ *     \arg @c 2 magnitude
  * - @b -n @e int
  *   - noise type
  *     \arg @c 0 none
@@ -110,11 +121,13 @@ void PrintUsage(std::ostream* stream) {
 int main(int argc, char* argv[]) {
   int frame_period(kDefaultFramePeriod);
   int interpolation_period(kDefaultInterpolationPeriod);
+  sptk::ExcitationGeneration::NormalizationType normalization_type(
+      kDefaultNormalizationType);
   NoiseType noise_type(kDefaultNoiseType);
   int seed(kDefaultSeed);
 
   for (;;) {
-    const int option_char(getopt_long(argc, argv, "p:i:n:s:h", NULL, NULL));
+    const int option_char(getopt_long(argc, argv, "p:i:N:n:s:h", NULL, NULL));
     if (-1 == option_char) break;
 
     switch (option_char) {
@@ -140,6 +153,25 @@ int main(int argc, char* argv[]) {
         }
         break;
       }
+      case 'N': {
+        const int min(0);
+        const int max(
+            static_cast<int>(sptk::ExcitationGeneration::NormalizationType::
+                                 kNumNormalizationTypes) -
+            1);
+        int tmp;
+        if (!sptk::ConvertStringToInteger(optarg, &tmp) ||
+            !sptk::IsInRange(tmp, min, max)) {
+          std::ostringstream error_message;
+          error_message << "The argument for the -N option must be an integer "
+                        << "in the range of " << min << " to " << max;
+          sptk::PrintErrorMessage("excite", error_message);
+          return 1;
+        }
+        normalization_type =
+            static_cast<sptk::ExcitationGeneration::NormalizationType>(tmp);
+        break;
+      }
       case 'n': {
         const int min(0);
         const int max(static_cast<int>(kNumNoiseTypes) - 1);
@@ -243,7 +275,8 @@ int main(int argc, char* argv[]) {
       }
     }
     sptk::ExcitationGeneration excitation_generation(
-        &input_source_interpolation_with_magic_number, random_generation);
+        &input_source_interpolation_with_magic_number, random_generation,
+        normalization_type);
 
     if (!excitation_generation.IsValid()) {
       std::ostringstream error_message;

src/main/frame.cc

@@ -20,6 +20,7 @@
 #include <iostream>   // std::cerr, std::cin, std::cout, std::endl, etc.
 #include <numeric>    // std::accumulate
 #include <sstream>    // std::ostringstream
+#include <string>     // std::char_traits
 #include <vector>     // std::vector
 
 #include "GETOPT/ya_getopt.h"
@@ -274,7 +275,7 @@ int main(int argc, char* argv[]) {
   // Extract the remaining frames.
   const int overlap(frame_length - frame_period);
   if (0 < overlap) {
-    bool is_eof(input_stream.peek() == std::ios::traits_type::eof());
+    bool is_eof(input_stream.peek() == std::char_traits<char>::eof());
     int last_data_position_in_frame(center + actual_read_size - 1);
     while (center <= last_data_position_in_frame) {
       if (is_eof) {
@@ -302,7 +303,7 @@ int main(int argc, char* argv[]) {
           return 1;
         }
 
-        if (input_stream.peek() == std::iostream::traits_type::eof()) {
+        if (input_stream.peek() == std::char_traits<char>::eof()) {
           last_data_position_in_frame = overlap + actual_read_size - 1;
           is_eof = true;
         }