porting the dithering function from kaldi

.gitignore

@@ -3,3 +3,6 @@ dist
 *.egg-info
 **/__pycache__
 build-arm-linux-gnueabihf
+
+# vim
+*.sw?

kaldi-native-fbank/csrc/CMakeLists.txt

@@ -5,6 +5,7 @@ set(sources
   feature-functions.cc
   feature-window.cc
   fftsg.cc
+  kaldi-math.cc
   mel-computations.cc
   online-feature.cc
   rfft.cc

kaldi-native-fbank/csrc/feature-window.cc

@@ -5,16 +5,13 @@
 // This file is copied/modified from kaldi/src/feat/feature-window.cc
 
 #include "kaldi-native-fbank/csrc/feature-window.h"
+#include "kaldi-native-fbank/csrc/kaldi-math.h"
 
 #include <algorithm>
 #include <cmath>
 #include <limits>
 #include <vector>
 
-#ifndef M_2PI
-#define M_2PI 6.283185307179586476925286766559005
-#endif
-
 namespace knf {
 
 std::ostream &operator<<(std::ostream &os, const FrameExtractionOptions &opts) {
@@ -205,6 +202,14 @@ float InnerProduct(const float *a, const float *b, int32_t n) {
   return sum;
 }
 
+void Dither(float *d, int32_t n, float dither_value) {
+  if (dither_value == 0.0)
+    return;
+  RandomState rstate;
+  for (int32_t i = 0; i < n; i++)
+    d[i] += RandGauss(&rstate) * dither_value;
+}
+
 static void Preemphasize(float *d, int32_t n, float preemph_coeff) {
   if (preemph_coeff == 0.0) {
     return;
@@ -223,6 +228,9 @@ void ProcessWindow(const FrameExtractionOptions &opts,
                    float *log_energy_pre_window /*= nullptr*/) {
   int32_t frame_length = opts.WindowSize();
 
+  if (opts.dither != 0.0)
+    Dither(window, frame_length, opts.dither);
+
   if (opts.remove_dc_offset) {
     RemoveDcOffset(window, frame_length);
   }

kaldi-native-fbank/csrc/feature-window.h

@@ -31,7 +31,8 @@ struct FrameExtractionOptions {
   float samp_freq = 16000;
   float frame_shift_ms = 10.0f;   // in milliseconds.
   float frame_length_ms = 25.0f;  // in milliseconds.
-  float dither = 1.0f;            // Amount of dithering, 0.0 means no dither.
+  float dither = 0.00003f;        // Amount of dithering, 0.0 means no dither.
+                                  // Value 0.00003f is equivalent to 1.0 in kaldi.
   float preemph_coeff = 0.97f;    // Preemphasis coefficient.
   bool remove_dc_offset = true;   // Subtract mean of wave before FFT.
   std::string window_type = "povey";  // e.g. Hamming window

kaldi-native-fbank/csrc/kaldi-math.cc

@@ -0,0 +1,38 @@
+// kaldi-native-fbank/csrc/kaldi-math.cc
+//
+// Copyright (c)  2024  Brno University of Technology (authors: Karel Vesely)
+
+// This file is an excerpt from kaldi/src/feat/kaldi-math.cc
+
+#include "kaldi-math.h"
+
+namespace knf {
+
+int Rand(struct RandomState* state) {
+#if !defined(_POSIX_THREAD_SAFE_FUNCTIONS)
+  // On Windows and Cygwin, just call Rand()
+  return rand();
+#else
+  if (state) {
+    return rand_r(&(state->seed));
+  } else {
+    std::lock_guard<std::mutex> lock(_RandMutex);
+    return rand();
+  }
+#endif
+}
+
+RandomState::RandomState() {
+  // we initialize it as Rand() + 27437 instead of just Rand(), because on some
+  // systems, e.g. at the very least Mac OSX Yosemite and later, it seems to be
+  // the case that rand_r when initialized with rand() will give you the exact
+  // same sequence of numbers that rand() will give if you keep calling rand()
+  // after that initial call.  This can cause problems with repeated sequences.
+  // For example if you initialize two RandomState structs one after the other
+  // without calling rand() in between, they would give you the same sequence
+  // offset by one (if we didn't have the "+ 27437" in the code).  27437 is just
+  // a randomly chosen prime number.
+  seed = unsigned(Rand()) + 27437;
+}
+
+}  // namespace knf

kaldi-native-fbank/csrc/kaldi-math.h

@@ -0,0 +1,43 @@
+// kaldi-native-fbank/csrc/kaldi-math.h
+//
+// Copyright (c)  2024  Brno University of Technology (authors: Karel Vesely)
+
+// This file is an excerpt from kaldi/src/feat/kaldi-math.h
+
+#pragma once
+
+#include <cmath>  // logf, sqrtf, cosf
+#include <cstdlib>  // RAND_MAX
+
+#ifndef M_PI
+#define M_PI 3.1415926535897932384626433832795
+#endif
+
+#ifndef M_2PI
+#define M_2PI 6.283185307179586476925286766559005
+#endif
+
+namespace knf {
+
+inline float Log(float x) { return logf(x); }
+
+// Returns a random integer between 0 and RAND_MAX, inclusive
+int Rand(struct RandomState* state = NULL);
+
+// State for thread-safe random number generator
+struct RandomState {
+  RandomState();
+  unsigned seed;
+};
+
+/// Returns a random number strictly between 0 and 1.
+inline float RandUniform(struct RandomState* state = NULL) {
+  return static_cast<float>((Rand(state) + 1.0) / (RAND_MAX+2.0));
+}
+
+inline float RandGauss(struct RandomState* state = NULL) {
+  return static_cast<float>(sqrtf (-2 * Log(RandUniform(state)))
+                            * cosf(2*M_PI*RandUniform(state)));
+}
+
+}  // namespace knf