removesilence: add threshold property

silence thresold was hardcoded to -60 dB, now it is configurable
using this new property

Closes #63

3 files changed, 38 insertions, 5 deletions

diff --git a/gst/removesilence/gstremovesilence.c b/gst/removesilence/gstremovesilence.c
index c861f29be..2debf5721 100644
--- a/gst/removesilence/gstremovesilence.c
+++ b/gst/removesilence/gstremovesilence.c
@@ -52,6 +52,7 @@ GST_DEBUG_CATEGORY_STATIC (gst_remove_silence_debug);
 #define MINIMUM_SILENCE_TIME_MIN  0
 #define MINIMUM_SILENCE_TIME_MAX  10000000000
 #define MINIMUM_SILENCE_TIME_DEF  0
+#define DEFAULT_VAD_THRESHOLD -60
 
 /* Filter signals and args */
 enum
@@ -65,6 +66,7 @@ enum
   PROP_0,
   PROP_REMOVE,
   PROP_HYSTERESIS,
+  PROP_THRESHOLD,
   PROP_SQUASH,
   PROP_SILENT,
   PROP_MINIMUM_SILENCE_BUFFERS,
@@ -132,6 +134,12 @@ gst_remove_silence_class_init (GstRemoveSilenceClass * klass)
           "Set the hysteresis (on samples) used on the internal VAD",
           1, G_MAXUINT64, DEFAULT_VAD_HYSTERESIS, G_PARAM_READWRITE));
 
+  g_object_class_install_property (gobject_class, PROP_THRESHOLD,
+      g_param_spec_int ("threshold",
+          "Threshold",
+          "Set the silence threshold used on the internal VAD in dB",
+          -70, 70, DEFAULT_VAD_THRESHOLD, G_PARAM_READWRITE));
+
   g_object_class_install_property (gobject_class, PROP_SQUASH,
       g_param_spec_boolean ("squash", "Squash",
           "Set to true to retimestamp buffers when silence is removed and so avoid timestamp gap",
@@ -180,7 +188,7 @@ gst_remove_silence_class_init (GstRemoveSilenceClass * klass)
 static void
 gst_remove_silence_init (GstRemoveSilence * filter)
 {
-  filter->vad = vad_new (DEFAULT_VAD_HYSTERESIS);
+  filter->vad = vad_new (DEFAULT_VAD_HYSTERESIS, DEFAULT_VAD_THRESHOLD);
   filter->remove = FALSE;
   filter->squash = FALSE;
   filter->ts_offset = 0;
@@ -221,6 +229,9 @@ gst_remove_silence_set_property (GObject * object, guint prop_id,
     case PROP_HYSTERESIS:
       vad_set_hysteresis (filter->vad, g_value_get_uint64 (value));
       break;
+    case PROP_THRESHOLD:
+      vad_set_threshold (filter->vad, g_value_get_int (value));
+      break;
     case PROP_SQUASH:
       filter->squash = g_value_get_boolean (value);
       break;
@@ -252,6 +263,9 @@ gst_remove_silence_get_property (GObject * object, guint prop_id,
     case PROP_HYSTERESIS:
       g_value_set_uint64 (value, vad_get_hysteresis (filter->vad));
       break;
+    case PROP_THRESHOLD:
+      g_value_set_int (value, vad_get_threshold_as_db (filter->vad));
+      break;
     case PROP_SQUASH:
       g_value_set_boolean (value, filter->squash);
       break;
diff --git a/gst/removesilence/vad_private.c b/gst/removesilence/vad_private.c
index 2c495967e..369e18e6f 100644
--- a/gst/removesilence/vad_private.c
+++ b/gst/removesilence/vad_private.c
@@ -22,11 +22,11 @@
 #include <string.h>
 #include <stdio.h>
 #include <stdlib.h>
+#include <math.h>
 #include <glib.h>
 #include "vad_private.h"
 
 #define VAD_POWER_ALPHA     0x0800      /* Q16 */
-#define VAD_POWER_THRESHOLD 0x000010C7  /* -60 dB (square wave) */
 #define VAD_ZCR_THRESHOLD   0
 #define VAD_BUFFER_SIZE     256
 
@@ -59,15 +59,17 @@ struct _vad_s
   guint64 hysteresis;
   guint64 vad_samples;
   guint64 vad_power;
+  guint64 threshold;
   long vad_zcr;
 };
 
 VADFilter *
-vad_new (guint64 hysteresis)
+vad_new (guint64 hysteresis, gint threshold)
 {
   VADFilter *vad = malloc (sizeof (VADFilter));
   vad_reset (vad);
   vad->hysteresis = hysteresis;
+  vad_set_threshold (vad, threshold);
   return vad;
 }
 
@@ -99,6 +101,19 @@ vad_get_hysteresis (struct _vad_s *p)
   return p->hysteresis;
 }
 
+void
+vad_set_threshold (struct _vad_s *p, gint threshold_db)
+{
+  gint power = (gint) (threshold_db / 10.0);
+  p->threshold = (guint64) (pow (10, (power)) * 4294967295);
+}
+
+gint
+vad_get_threshold_as_db (struct _vad_s *p)
+{
+  return (gint) (10 * log10 (p->threshold / 4294967295.0));
+}
+
 gint
 vad_update (struct _vad_s * p, gint16 * data, gint len)
 {
@@ -129,7 +144,7 @@ vad_update (struct _vad_s * p, gint16 * data, gint len)
         ((sample & 0x8000) != (p->cqueue.base.s[tail] & 0x8000)) ? 1 : -1;
   }
 
-  frame_type = (p->vad_power > VAD_POWER_THRESHOLD
+  frame_type = (p->vad_power > p->threshold
       && p->vad_zcr < VAD_ZCR_THRESHOLD) ? VAD_VOICE : VAD_SILENCE;
 
   if (p->vad_state != frame_type) {
diff --git a/gst/removesilence/vad_private.h b/gst/removesilence/vad_private.h
index 08eef627a..530a38c5e 100644
--- a/gst/removesilence/vad_private.h
+++ b/gst/removesilence/vad_private.h
@@ -35,7 +35,11 @@ void vad_set_hysteresis(VADFilter *p, guint64 hysteresis);
 
 guint64 vad_get_hysteresis(VADFilter *p);
 
-VADFilter* vad_new(guint64 hysteresis);
+void vad_set_threshold(VADFilter *p, gint threshold_db);
+
+gint vad_get_threshold_as_db(VADFilter *p);
+
+VADFilter* vad_new(guint64 hysteresis, gint threshold);
 
 void vad_reset(VADFilter *p);