gst/audiofx/: Add simple voice removal element. Yay karaoke.
authorWim Taymans <wim.taymans@gmail.com>
Mon, 26 May 2008 15:51:41 +0000 (15:51 +0000)
committerWim Taymans <wim.taymans@gmail.com>
Mon, 26 May 2008 15:51:41 +0000 (15:51 +0000)
Original commit message from CVS:
* gst/audiofx/Makefile.am:
* gst/audiofx/audiofx.c: (plugin_init):
* gst/audiofx/audiovoice.c: (gst_audio_voice_base_init),
(gst_audio_voice_class_init), (gst_audio_voice_init),
(update_filter), (gst_audio_voice_set_property),
(gst_audio_voice_get_property), (gst_audio_voice_setup),
(gst_audio_voice_transform_int), (gst_audio_voice_transform_float),
(gst_audio_voice_transform_ip):
* gst/audiofx/audiovoice.h:
Add simple voice removal element. Yay karaoke.

ChangeLog
gst/audiofx/Makefile.am
gst/audiofx/audiofx.c
gst/audiofx/audiovoice.c [new file with mode: 0644]
gst/audiofx/audiovoice.h [new file with mode: 0644]

index 31c7fe7..a342b60 100644 (file)
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,5 +1,18 @@
 2008-05-26  Wim Taymans  <wim.taymans@collabora.co.uk>
 
+       * gst/audiofx/Makefile.am:
+       * gst/audiofx/audiofx.c: (plugin_init):
+       * gst/audiofx/audiovoice.c: (gst_audio_voice_base_init),
+       (gst_audio_voice_class_init), (gst_audio_voice_init),
+       (update_filter), (gst_audio_voice_set_property),
+       (gst_audio_voice_get_property), (gst_audio_voice_setup),
+       (gst_audio_voice_transform_int), (gst_audio_voice_transform_float),
+       (gst_audio_voice_transform_ip):
+       * gst/audiofx/audiovoice.h:
+       Add simple voice removal element. Yay karaoke.
+
+2008-05-26  Wim Taymans  <wim.taymans@collabora.co.uk>
+
        Patch by: William M. Brack <wbrack at mmm dot com dot hk>
 
        * sys/v4l2/v4l2src_calls.c: (gst_v4l2src_probe_caps_for_format):
index 3c0df21..28de5e2 100644 (file)
@@ -8,6 +8,7 @@ libgstaudiofx_la_SOURCES = audiofx.c\
        audioinvert.c \
        audioamplify.c \
        audiodynamic.c \
+       audiovoice.c \
        audiocheblimit.c \
        audiochebband.c \
        audiowsincband.c \
@@ -31,6 +32,7 @@ noinst_HEADERS = audiopanorama.h \
        audioinvert.h \
        audioamplify.h \
        audiodynamic.h \
+       audiovoice.h \
        audiocheblimit.h \
        audiochebband.h \
        audiowsincband.h \
index ea6a8c2..f9b62ca 100644 (file)
@@ -27,6 +27,7 @@
 
 #include "audiopanorama.h"
 #include "audioinvert.h"
+#include "audiovoice.h"
 #include "audioamplify.h"
 #include "audiodynamic.h"
 #include "audiocheblimit.h"
@@ -49,6 +50,8 @@ plugin_init (GstPlugin * plugin)
           GST_TYPE_AUDIO_PANORAMA) &&
       gst_element_register (plugin, "audioinvert", GST_RANK_NONE,
           GST_TYPE_AUDIO_INVERT) &&
+      gst_element_register (plugin, "audiovoice", GST_RANK_NONE,
+          GST_TYPE_AUDIO_VOICE) &&
       gst_element_register (plugin, "audioamplify", GST_RANK_NONE,
           GST_TYPE_AUDIO_AMPLIFY) &&
       gst_element_register (plugin, "audiodynamic", GST_RANK_NONE,
diff --git a/gst/audiofx/audiovoice.c b/gst/audiofx/audiovoice.c
new file mode 100644 (file)
index 0000000..08916c1
--- /dev/null
@@ -0,0 +1,359 @@
+/* 
+ * GStreamer
+ * Copyright (C) 2008 Wim Taymans <wim.taymans@gmail.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Library General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Library General Public License for more details.
+ *
+ * You should have received a copy of the GNU Library General Public
+ * License along with this library; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 02111-1307, USA.
+ */
+
+/**
+ * SECTION:element-audiovoice
+ * @short_description: Voice removal element
+ *
+ * <refsect2>
+ * Remove the voice from audio by removing the center channel.
+ * This plugin is useful for karaoke applications.
+ * <title>Example launch line</title>
+ * <para>
+ * <programlisting>
+ * gst-launch filesrc location="song.ogg" ! oggdemux ! vorbisdec ! audiovoice ! audioconvert ! alsasink
+ * </programlisting>
+ * </para>
+ * </refsect2>
+ */
+
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#include <math.h>
+
+#include <gst/gst.h>
+#include <gst/base/gstbasetransform.h>
+#include <gst/audio/audio.h>
+#include <gst/audio/gstaudiofilter.h>
+#include <gst/controller/gstcontroller.h>
+
+#include "audiovoice.h"
+
+#define GST_CAT_DEFAULT gst_audio_voice_debug
+GST_DEBUG_CATEGORY_STATIC (GST_CAT_DEFAULT);
+
+static const GstElementDetails element_details =
+GST_ELEMENT_DETAILS ("AudioVoice",
+    "Filter/Effect/Audio",
+    "Removes voice from sound",
+    "Wim Taymans <wim.taymans@gmail.com>");
+
+/* Filter signals and args */
+enum
+{
+  /* FILL ME */
+  LAST_SIGNAL
+};
+
+#define DEFAULT_LEVEL          1.0
+#define DEFAULT_MONO_LEVEL     1.0
+#define DEFAULT_FILTER_BAND    220.0
+#define DEFAULT_FILTER_WIDTH   100.0
+
+enum
+{
+  PROP_0,
+  PROP_LEVEL,
+  PROP_MONO_LEVEL,
+  PROP_FILTER_BAND,
+  PROP_FILTER_WIDTH,
+  PROP_LAST
+};
+
+#define ALLOWED_CAPS \
+    "audio/x-raw-int,"                                                \
+    " depth=(int)16,"                                                 \
+    " width=(int)16,"                                                 \
+    " endianness=(int)BYTE_ORDER,"                                    \
+    " signed=(bool)TRUE,"                                             \
+    " rate=(int)[1,MAX],"                                             \
+    " channels=(int)[1,MAX]; "                                        \
+    "audio/x-raw-float,"                                              \
+    " width=(int)32,"                                                 \
+    " endianness=(int)BYTE_ORDER,"                                    \
+    " rate=(int)[1,MAX],"                                             \
+    " channels=(int)[1,MAX]"
+
+#define DEBUG_INIT(bla) \
+  GST_DEBUG_CATEGORY_INIT (gst_audio_voice_debug, "audiovoice", 0, "audiovoice element");
+
+GST_BOILERPLATE_FULL (GstAudioVoice, gst_audio_voice, GstAudioFilter,
+    GST_TYPE_AUDIO_FILTER, DEBUG_INIT);
+
+static void gst_audio_voice_set_property (GObject * object, guint prop_id,
+    const GValue * value, GParamSpec * pspec);
+static void gst_audio_voice_get_property (GObject * object, guint prop_id,
+    GValue * value, GParamSpec * pspec);
+
+static gboolean gst_audio_voice_setup (GstAudioFilter * filter,
+    GstRingBufferSpec * format);
+static GstFlowReturn gst_audio_voice_transform_ip (GstBaseTransform * base,
+    GstBuffer * buf);
+
+static void gst_audio_voice_transform_int (GstAudioVoice * filter,
+    gint16 * data, guint num_samples);
+static void gst_audio_voice_transform_float (GstAudioVoice * filter,
+    gfloat * data, guint num_samples);
+
+/* GObject vmethod implementations */
+
+static void
+gst_audio_voice_base_init (gpointer klass)
+{
+  GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
+  GstCaps *caps;
+
+  gst_element_class_set_details (element_class, &element_details);
+
+  caps = gst_caps_from_string (ALLOWED_CAPS);
+  gst_audio_filter_class_add_pad_templates (GST_AUDIO_FILTER_CLASS (klass),
+      caps);
+  gst_caps_unref (caps);
+}
+
+static void
+gst_audio_voice_class_init (GstAudioVoiceClass * klass)
+{
+  GObjectClass *gobject_class;
+
+  gobject_class = (GObjectClass *) klass;
+  gobject_class->set_property = gst_audio_voice_set_property;
+  gobject_class->get_property = gst_audio_voice_get_property;
+
+  g_object_class_install_property (gobject_class, PROP_LEVEL,
+      g_param_spec_float ("level", "Level",
+          "Level of the effect (1.0 = full)", 0.0, 1.0, DEFAULT_LEVEL,
+          G_PARAM_READWRITE | GST_PARAM_CONTROLLABLE));
+
+  g_object_class_install_property (gobject_class, PROP_MONO_LEVEL,
+      g_param_spec_float ("mono-level", "Mono Level",
+          "Level of the mono channel (1.0 = full)", 0.0, 1.0, DEFAULT_LEVEL,
+          G_PARAM_READWRITE | GST_PARAM_CONTROLLABLE));
+
+  g_object_class_install_property (gobject_class, PROP_FILTER_BAND,
+      g_param_spec_float ("filter-band", "Filter Band",
+          "The Frequency band of the filter", 0.0, 441.0, DEFAULT_FILTER_BAND,
+          G_PARAM_READWRITE | GST_PARAM_CONTROLLABLE));
+
+  g_object_class_install_property (gobject_class, PROP_FILTER_WIDTH,
+      g_param_spec_float ("filter-width", "Filter Width",
+          "The Frequency width of the filter", 0.0, 100.0, DEFAULT_FILTER_WIDTH,
+          G_PARAM_READWRITE | GST_PARAM_CONTROLLABLE));
+
+  GST_AUDIO_FILTER_CLASS (klass)->setup =
+      GST_DEBUG_FUNCPTR (gst_audio_voice_setup);
+  GST_BASE_TRANSFORM_CLASS (klass)->transform_ip =
+      GST_DEBUG_FUNCPTR (gst_audio_voice_transform_ip);
+}
+
+static void
+gst_audio_voice_init (GstAudioVoice * filter, GstAudioVoiceClass * klass)
+{
+  gst_base_transform_set_in_place (GST_BASE_TRANSFORM (filter), TRUE);
+  gst_base_transform_set_gap_aware (GST_BASE_TRANSFORM (filter), TRUE);
+
+  filter->level = DEFAULT_LEVEL;
+  filter->mono_level = DEFAULT_MONO_LEVEL;
+  filter->filter_band = DEFAULT_FILTER_BAND;
+  filter->filter_width = DEFAULT_FILTER_WIDTH;
+}
+
+static void
+update_filter (GstAudioVoice * filter, gint rate)
+{
+  gfloat A, B, C;
+
+  if (rate == 0)
+    return;
+
+  C = exp (-2 * M_PI * filter->filter_width / rate);
+  B = -4 * C / (1 + C) * cos (2 * M_PI * filter->filter_band / rate);
+  A = sqrt (1 - B * B / (4 * C)) * (1 - C);
+
+  filter->A = A;
+  filter->B = B;
+  filter->C = C;
+  filter->y1 = 0.0;
+  filter->y2 = 0.0;
+}
+
+static void
+gst_audio_voice_set_property (GObject * object, guint prop_id,
+    const GValue * value, GParamSpec * pspec)
+{
+  GstAudioVoice *filter;
+
+  filter = GST_AUDIO_VOICE (object);
+
+  switch (prop_id) {
+    case PROP_LEVEL:
+      filter->level = g_value_get_float (value);
+      break;
+    case PROP_MONO_LEVEL:
+      filter->mono_level = g_value_get_float (value);
+      break;
+    case PROP_FILTER_BAND:
+      filter->filter_band = g_value_get_float (value);
+      update_filter (filter, filter->rate);
+      break;
+    case PROP_FILTER_WIDTH:
+      filter->filter_width = g_value_get_float (value);
+      update_filter (filter, filter->rate);
+      break;
+    default:
+      G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
+      break;
+  }
+}
+
+static void
+gst_audio_voice_get_property (GObject * object, guint prop_id,
+    GValue * value, GParamSpec * pspec)
+{
+  GstAudioVoice *filter;
+
+  filter = GST_AUDIO_VOICE (object);
+
+  switch (prop_id) {
+    case PROP_LEVEL:
+      g_value_set_float (value, filter->level);
+      break;
+    case PROP_MONO_LEVEL:
+      g_value_set_float (value, filter->mono_level);
+      break;
+    case PROP_FILTER_BAND:
+      g_value_set_float (value, filter->filter_band);
+      break;
+    case PROP_FILTER_WIDTH:
+      g_value_set_float (value, filter->filter_width);
+      break;
+    default:
+      G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
+      break;
+  }
+}
+
+/* GstAudioFilter vmethod implementations */
+
+static gboolean
+gst_audio_voice_setup (GstAudioFilter * base, GstRingBufferSpec * format)
+{
+  GstAudioVoice *filter = GST_AUDIO_VOICE (base);
+  gboolean ret = TRUE;
+
+  filter->channels = format->channels;
+  filter->rate = format->rate;
+
+  if (format->type == GST_BUFTYPE_FLOAT && format->width == 32)
+    filter->process = (GstAudioVoiceProcessFunc)
+        gst_audio_voice_transform_float;
+  else if (format->type == GST_BUFTYPE_LINEAR && format->width == 16)
+    filter->process = (GstAudioVoiceProcessFunc)
+        gst_audio_voice_transform_int;
+  else
+    ret = FALSE;
+
+  update_filter (filter, format->rate);
+
+  return ret;
+}
+
+static void
+gst_audio_voice_transform_int (GstAudioVoice * filter,
+    gint16 * data, guint num_samples)
+{
+  gint i, l, r, o, x;
+  gint channels;
+  gdouble y;
+  gint level;
+
+  channels = filter->channels;
+  level = filter->level * 256;
+
+  for (i = 0; i < num_samples; i += channels) {
+    /* get left and right inputs */
+    l = data[i];
+    r = data[i + 1];
+    /* do filtering */
+    x = (l + r) / 2;
+    y = (filter->A * x - filter->B * filter->y1) - filter->C * filter->y2;
+    filter->y2 = filter->y1;
+    filter->y1 = y;
+    /* filter mono signal */
+    o = (int) (y * filter->mono_level);
+    o = CLAMP (o, G_MININT16, G_MAXINT16);
+    o = (o * level) >> 8;
+    /* now cut the center */
+    x = l - ((r * level) >> 8) + o;
+    r = r - ((l * level) >> 8) + o;
+    data[i] = CLAMP (x, G_MININT16, G_MAXINT16);
+    data[i + 1] = CLAMP (r, G_MININT16, G_MAXINT16);
+  }
+}
+
+static void
+gst_audio_voice_transform_float (GstAudioVoice * filter,
+    gfloat * data, guint num_samples)
+{
+  gint i;
+  gint channels;
+  gdouble l, r, o;
+  gdouble y;
+
+  channels = filter->channels;
+
+  for (i = 0; i < num_samples; i += channels) {
+    /* get left and right inputs */
+    l = data[i];
+    r = data[i + 1];
+    /* do filtering */
+    y = (filter->A * ((l + r) / 2.0) - filter->B * filter->y1) -
+        filter->C * filter->y2;
+    filter->y2 = filter->y1;
+    filter->y1 = y;
+    /* filter mono signal */
+    o = y * filter->mono_level * filter->level;
+    /* now cut the center */
+    data[i] = l - (r * filter->level) + o;
+    data[i + 1] = r - (l * filter->level) + o;
+  }
+}
+
+/* GstBaseTransform vmethod implementations */
+static GstFlowReturn
+gst_audio_voice_transform_ip (GstBaseTransform * base, GstBuffer * buf)
+{
+  GstAudioVoice *filter = GST_AUDIO_VOICE (base);
+  guint num_samples =
+      GST_BUFFER_SIZE (buf) / (GST_AUDIO_FILTER (filter)->format.width / 8);
+
+  if (GST_CLOCK_TIME_IS_VALID (GST_BUFFER_TIMESTAMP (buf)))
+    gst_object_sync_values (G_OBJECT (filter), GST_BUFFER_TIMESTAMP (buf));
+
+  if (gst_base_transform_is_passthrough (base) ||
+      G_UNLIKELY (GST_BUFFER_FLAG_IS_SET (buf, GST_BUFFER_FLAG_GAP)))
+    return GST_FLOW_OK;
+
+  filter->process (filter, GST_BUFFER_DATA (buf), num_samples);
+
+  return GST_FLOW_OK;
+}
diff --git a/gst/audiofx/audiovoice.h b/gst/audiofx/audiovoice.h
new file mode 100644 (file)
index 0000000..cf3ff4f
--- /dev/null
@@ -0,0 +1,70 @@
+/* 
+ * GStreamer
+ * Copyright (C) 2008 Wim Taymans <wim.taymans@gmail.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Library General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Library General Public License for more details.
+ *
+ * You should have received a copy of the GNU Library General Public
+ * License along with this library; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 02111-1307, USA.
+ */
+
+#ifndef __GST_AUDIO_VOICE_H__
+#define __GST_AUDIO_VOICE_H__
+
+#include <gst/gst.h>
+#include <gst/base/gstbasetransform.h>
+#include <gst/audio/audio.h>
+#include <gst/audio/gstaudiofilter.h>
+
+G_BEGIN_DECLS
+#define GST_TYPE_AUDIO_VOICE            (gst_audio_voice_get_type())
+#define GST_AUDIO_VOICE(obj)            (G_TYPE_CHECK_INSTANCE_CAST((obj),GST_TYPE_AUDIO_VOICE,GstAudioVoice))
+#define GST_IS_AUDIO_VOICE(obj)         (G_TYPE_CHECK_INSTANCE_TYPE((obj),GST_TYPE_AUDIO_VOICE))
+#define GST_AUDIO_VOICE_CLASS(klass)    (G_TYPE_CHECK_CLASS_CAST((klass) ,GST_TYPE_AUDIO_VOICE,GstAudioVoiceClass))
+#define GST_IS_AUDIO_VOICE_CLASS(klass) (G_TYPE_CHECK_CLASS_TYPE((klass) ,GST_TYPE_AUDIO_VOICE))
+#define GST_AUDIO_VOICE_GET_CLASS(obj)  (G_TYPE_INSTANCE_GET_CLASS((obj) ,GST_TYPE_AUDIO_VOICE,GstAudioVoiceClass))
+typedef struct _GstAudioVoice GstAudioVoice;
+typedef struct _GstAudioVoiceClass GstAudioVoiceClass;
+
+typedef void (*GstAudioVoiceProcessFunc) (GstAudioVoice *, guint8 *, guint);
+
+struct _GstAudioVoice
+{
+  GstAudioFilter audiofilter;
+
+  gint channels;
+  gint rate;
+
+  /* properties */
+  gfloat level;
+  gfloat mono_level;
+  gfloat filter_band;
+  gfloat filter_width;
+
+  /* filter coef */
+  gfloat A, B, C;
+  gfloat y1, y2;
+
+  /* < private > */
+  GstAudioVoiceProcessFunc process;
+};
+
+struct _GstAudioVoiceClass
+{
+  GstAudioFilterClass parent;
+};
+
+GType gst_audio_voice_get_type (void);
+
+G_END_DECLS
+#endif /* __GST_AUDIO_VOICE_H__ */