1 /* GStreamer
2 * Copyright (C) 2011 Tiago Katcipis <tiagokatcipis@gmail.com>
3 * Copyright (C) 2011 Paulo Pizarro <paulo.pizarro@gmail.com>
4 * Copyright (C) 2012-2016 Nicola Murino <nicola.murino@gmail.com>
5 *
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Library General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
10 *
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Library General Public License for more details.
15 *
16 * You should have received a copy of the GNU Library General Public
17 * License along with this library; if not, write to the
18 * Free Software Foundation, Inc., 51 Franklin St, Fifth Floor,
19 * Boston, MA 02110-1301, USA.
20 */
21
22 /**
23 * SECTION:element-removesilence
24 * @title: removesilence
25 *
26 * Removes all silence periods from an audio stream, dropping silence buffers.
27 * If the "silent" property is disabled, removesilence will generate
28 * bus messages named "removesilence".
29 * The message's structure contains one of these fields:
30 *
31 * - #guint64 "silence_detected": the PTS for the first silent buffer after a non silence period.
32 *
33 * - #guint64 "silence_finished": the PTS for the first non silent buffer after a silence period.
34 *
35 * ## Example launch line
36 * |[
37 * gst-launch-1.0 -v -m filesrc location="audiofile" ! decodebin ! removesilence remove=true ! wavenc ! filesink location=without_audio.wav
38 * ]|
39 *
40 */
41
42 #ifdef HAVE_CONFIG_H
43 #include "config.h"
44 #endif
45
46 #include <gst/gst.h>
47 #include <gst/base/gstbasetransform.h>
48 #include <gst/audio/audio.h>
49
50 #include "gstremovesilence.h"
51
52
53 GST_DEBUG_CATEGORY_STATIC (gst_remove_silence_debug);
54 #define GST_CAT_DEFAULT gst_remove_silence_debug
55 #define DEFAULT_VAD_HYSTERESIS 480 /* 60 mseg */
56 #define MINIMUM_SILENCE_BUFFERS_MIN 0
57 #define MINIMUM_SILENCE_BUFFERS_MAX 10000
58 #define MINIMUM_SILENCE_BUFFERS_DEF 0
59 #define MINIMUM_SILENCE_TIME_MIN 0
60 #define MINIMUM_SILENCE_TIME_MAX 10000000000
61 #define MINIMUM_SILENCE_TIME_DEF 0
62 #define DEFAULT_VAD_THRESHOLD -60
63
64 /* Filter signals and args */
65 enum
66 {
67 /* FILL ME */
68 LAST_SIGNAL
69 };
70
71 enum
72 {
73 PROP_0,
74 PROP_REMOVE,
75 PROP_HYSTERESIS,
76 PROP_THRESHOLD,
77 PROP_SQUASH,
78 PROP_SILENT,
79 PROP_MINIMUM_SILENCE_BUFFERS,
80 PROP_MINIMUM_SILENCE_TIME
81 };
82
83
84 static GstStaticPadTemplate sink_template = GST_STATIC_PAD_TEMPLATE ("sink",
85 GST_PAD_SINK,
86 GST_PAD_ALWAYS,
87 GST_STATIC_CAPS ("audio/x-raw, "
88 "format = (string) " GST_AUDIO_NE (S16) ", "
89 "layout = (string) interleaved, "
90 "rate = (int) [ 1, MAX ], " "channels = (int) 1"));
91
92 static GstStaticPadTemplate src_template = GST_STATIC_PAD_TEMPLATE ("src",
93 GST_PAD_SRC,
94 GST_PAD_ALWAYS,
95 GST_STATIC_CAPS ("audio/x-raw, "
96 "format = (string) " GST_AUDIO_NE (S16) ", "
97 "layout = (string) interleaved, "
98 "rate = (int) [ 1, MAX ], " "channels = (int) 1"));
99
100
101 #define DEBUG_INIT(bla) \
102 GST_DEBUG_CATEGORY_INIT (gst_remove_silence_debug, "removesilence", 0, "removesilence element")
103
104 #define gst_remove_silence_parent_class parent_class
105 G_DEFINE_TYPE_WITH_CODE (GstRemoveSilence, gst_remove_silence,
106 GST_TYPE_BASE_TRANSFORM, DEBUG_INIT (0));
107
108 static void gst_remove_silence_set_property (GObject * object, guint prop_id,
109 const GValue * value, GParamSpec * pspec);
110 static void gst_remove_silence_get_property (GObject * object, guint prop_id,
111 GValue * value, GParamSpec * pspec);
112
113 static gboolean gst_remove_silence_start (GstBaseTransform * trans);
114 static gboolean gst_remove_silence_sink_event (GstBaseTransform * trans,
115 GstEvent * event);
116 static GstFlowReturn gst_remove_silence_transform_ip (GstBaseTransform * base,
117 GstBuffer * buf);
118 static void gst_remove_silence_finalize (GObject * obj);
119
120 /* GObject vmethod implementations */
121
122 /* initialize the removesilence's class */
123 static void
gst_remove_silence_class_init(GstRemoveSilenceClass * klass)124 gst_remove_silence_class_init (GstRemoveSilenceClass * klass)
125 {
126 GObjectClass *gobject_class;
127 GstElementClass *gstelement_class;
128 GstBaseTransformClass *base_transform_class;
129
130 gobject_class = (GObjectClass *) klass;
131 gstelement_class = (GstElementClass *) klass;
132 base_transform_class = GST_BASE_TRANSFORM_CLASS (klass);
133
134 gobject_class->finalize = gst_remove_silence_finalize;
135 gobject_class->set_property = gst_remove_silence_set_property;
136 gobject_class->get_property = gst_remove_silence_get_property;
137
138 g_object_class_install_property (gobject_class, PROP_REMOVE,
139 g_param_spec_boolean ("remove", "Remove",
140 "Set to true to remove silence from the stream, false otherwhise",
141 FALSE, G_PARAM_READWRITE));
142
143 g_object_class_install_property (gobject_class, PROP_HYSTERESIS,
144 g_param_spec_uint64 ("hysteresis",
145 "Hysteresis",
146 "Set the hysteresis (on samples) used on the internal VAD",
147 1, G_MAXUINT64, DEFAULT_VAD_HYSTERESIS, G_PARAM_READWRITE));
148
149 g_object_class_install_property (gobject_class, PROP_THRESHOLD,
150 g_param_spec_int ("threshold",
151 "Threshold",
152 "Set the silence threshold used on the internal VAD in dB",
153 -70, 70, DEFAULT_VAD_THRESHOLD, G_PARAM_READWRITE));
154
155 g_object_class_install_property (gobject_class, PROP_SQUASH,
156 g_param_spec_boolean ("squash", "Squash",
157 "Set to true to retimestamp buffers when silence is removed and so avoid timestamp gap",
158 FALSE, G_PARAM_READWRITE));
159
160 g_object_class_install_property (gobject_class, PROP_SILENT,
161 g_param_spec_boolean ("silent", "Silent",
162 "Disable/enable bus message notifications for silence detected/finished",
163 TRUE, G_PARAM_READWRITE));
164
165 g_object_class_install_property (gobject_class, PROP_MINIMUM_SILENCE_BUFFERS,
166 g_param_spec_uint ("minimum-silence-buffers", "Minimum silence buffers",
167 "Define the minimum number of consecutive silence buffers before "
168 "removing silence, 0 means disabled. This will not introduce latency",
169 MINIMUM_SILENCE_BUFFERS_MIN, MINIMUM_SILENCE_BUFFERS_MAX,
170 MINIMUM_SILENCE_BUFFERS_DEF, G_PARAM_READWRITE));
171
172 g_object_class_install_property (gobject_class, PROP_MINIMUM_SILENCE_TIME,
173 g_param_spec_uint64 ("minimum_silence_time",
174 "Minimum silence time",
175 "Define the minimum silence time in nanoseconds before removing "
176 " silence, 0 means disabled. This will not introduce latency",
177 MINIMUM_SILENCE_TIME_MIN, MINIMUM_SILENCE_TIME_MAX,
178 MINIMUM_SILENCE_TIME_DEF, G_PARAM_READWRITE));
179
180 gst_element_class_set_static_metadata (gstelement_class,
181 "RemoveSilence",
182 "Filter/Effect/Audio",
183 "Removes all the silence periods from the audio stream.",
184 "Tiago Katcipis <tiagokatcipis@gmail.com>\n \
185 Paulo Pizarro <paulo.pizarro@gmail.com>\n \
186 Nicola Murino <nicola.murino@gmail.com>");
187
188 gst_element_class_add_static_pad_template (gstelement_class, &src_template);
189 gst_element_class_add_static_pad_template (gstelement_class, &sink_template);
190
191 base_transform_class->start = GST_DEBUG_FUNCPTR (gst_remove_silence_start);
192 base_transform_class->sink_event =
193 GST_DEBUG_FUNCPTR (gst_remove_silence_sink_event);
194 base_transform_class->transform_ip =
195 GST_DEBUG_FUNCPTR (gst_remove_silence_transform_ip);
196 }
197
198 static void
gst_remove_silence_reset(GstRemoveSilence * filter)199 gst_remove_silence_reset (GstRemoveSilence * filter)
200 {
201 filter->ts_offset = 0;
202 filter->silence_detected = FALSE;
203 filter->consecutive_silence_buffers = 0;
204 filter->consecutive_silence_time = 0;
205 }
206
207 /* initialize the new element
208 * instantiate pads and add them to element
209 * set pad calback functions
210 * initialize instance structure
211 */
212 static void
gst_remove_silence_init(GstRemoveSilence * filter)213 gst_remove_silence_init (GstRemoveSilence * filter)
214 {
215 filter->vad = vad_new (DEFAULT_VAD_HYSTERESIS, DEFAULT_VAD_THRESHOLD);
216 filter->remove = FALSE;
217 filter->squash = FALSE;
218 filter->silent = TRUE;
219 filter->minimum_silence_buffers = MINIMUM_SILENCE_BUFFERS_DEF;
220 filter->minimum_silence_time = MINIMUM_SILENCE_TIME_DEF;
221
222 gst_remove_silence_reset (filter);
223
224 if (!filter->vad) {
225 GST_DEBUG ("Error initializing VAD !!");
226 return;
227 }
228 }
229
230 static gboolean
gst_remove_silence_start(GstBaseTransform * trans)231 gst_remove_silence_start (GstBaseTransform * trans)
232 {
233 GstRemoveSilence *filter = GST_REMOVE_SILENCE (trans);
234
235 GST_INFO ("reset filter on start");
236 gst_remove_silence_reset (filter);
237
238 return TRUE;
239 }
240
241 static gboolean
gst_remove_silence_sink_event(GstBaseTransform * trans,GstEvent * event)242 gst_remove_silence_sink_event (GstBaseTransform * trans, GstEvent * event)
243 {
244 GstRemoveSilence *filter = GST_REMOVE_SILENCE (trans);
245
246 if (event->type == GST_EVENT_SEGMENT) {
247 GST_INFO ("reset filter on segment event");
248 gst_remove_silence_reset (filter);
249 }
250
251 return
252 GST_BASE_TRANSFORM_CLASS (gst_remove_silence_parent_class)->sink_event
253 (trans, event);
254 }
255
256 static void
gst_remove_silence_finalize(GObject * obj)257 gst_remove_silence_finalize (GObject * obj)
258 {
259 GstRemoveSilence *filter = GST_REMOVE_SILENCE (obj);
260 GST_DEBUG ("Destroying VAD");
261 vad_destroy (filter->vad);
262 filter->vad = NULL;
263 GST_DEBUG ("VAD Destroyed");
264 G_OBJECT_CLASS (parent_class)->finalize (obj);
265 }
266
267 static void
gst_remove_silence_set_property(GObject * object,guint prop_id,const GValue * value,GParamSpec * pspec)268 gst_remove_silence_set_property (GObject * object, guint prop_id,
269 const GValue * value, GParamSpec * pspec)
270 {
271 GstRemoveSilence *filter = GST_REMOVE_SILENCE (object);
272
273 switch (prop_id) {
274 case PROP_REMOVE:
275 filter->remove = g_value_get_boolean (value);
276 break;
277 case PROP_HYSTERESIS:
278 vad_set_hysteresis (filter->vad, g_value_get_uint64 (value));
279 break;
280 case PROP_THRESHOLD:
281 vad_set_threshold (filter->vad, g_value_get_int (value));
282 break;
283 case PROP_SQUASH:
284 filter->squash = g_value_get_boolean (value);
285 break;
286 case PROP_SILENT:
287 filter->silent = g_value_get_boolean (value);
288 break;
289 case PROP_MINIMUM_SILENCE_BUFFERS:
290 filter->minimum_silence_buffers = g_value_get_uint (value);
291 break;
292 case PROP_MINIMUM_SILENCE_TIME:
293 filter->minimum_silence_time = g_value_get_uint64 (value);
294 break;
295 default:
296 G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
297 break;
298 }
299 }
300
301 static void
gst_remove_silence_get_property(GObject * object,guint prop_id,GValue * value,GParamSpec * pspec)302 gst_remove_silence_get_property (GObject * object, guint prop_id,
303 GValue * value, GParamSpec * pspec)
304 {
305 GstRemoveSilence *filter = GST_REMOVE_SILENCE (object);
306
307 switch (prop_id) {
308 case PROP_REMOVE:
309 g_value_set_boolean (value, filter->remove);
310 break;
311 case PROP_HYSTERESIS:
312 g_value_set_uint64 (value, vad_get_hysteresis (filter->vad));
313 break;
314 case PROP_THRESHOLD:
315 g_value_set_int (value, vad_get_threshold_as_db (filter->vad));
316 break;
317 case PROP_SQUASH:
318 g_value_set_boolean (value, filter->squash);
319 break;
320 case PROP_SILENT:
321 g_value_set_boolean (value, filter->silent);
322 break;
323 case PROP_MINIMUM_SILENCE_BUFFERS:
324 g_value_set_uint (value, filter->minimum_silence_buffers);
325 break;
326 case PROP_MINIMUM_SILENCE_TIME:
327 g_value_set_uint64 (value, filter->minimum_silence_time);
328 break;
329 default:
330 G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
331 break;
332 }
333 }
334
335 static GstFlowReturn
gst_remove_silence_transform_ip(GstBaseTransform * trans,GstBuffer * inbuf)336 gst_remove_silence_transform_ip (GstBaseTransform * trans, GstBuffer * inbuf)
337 {
338 GstRemoveSilence *filter = NULL;
339 int frame_type;
340 GstMapInfo map;
341 gboolean consecutive_silence_reached;
342
343 filter = GST_REMOVE_SILENCE (trans);
344
345 gst_buffer_map (inbuf, &map, GST_MAP_READ);
346 frame_type =
347 vad_update (filter->vad, (gint16 *) map.data, map.size / sizeof (gint16));
348 gst_buffer_unmap (inbuf, &map);
349
350 if (frame_type == VAD_SILENCE) {
351 GST_DEBUG ("Silence detected");
352 filter->consecutive_silence_buffers++;
353 if (GST_BUFFER_DURATION_IS_VALID (inbuf)) {
354 filter->consecutive_silence_time += inbuf->duration;
355 } else {
356 GST_WARNING
357 ("Invalid buffer duration, consecutive_silence_time update not possible");
358 }
359 if (filter->minimum_silence_buffers == 0
360 && filter->minimum_silence_time == 0) {
361 consecutive_silence_reached = TRUE;
362 } else {
363 consecutive_silence_reached =
364 (filter->minimum_silence_buffers > 0
365 && filter->consecutive_silence_buffers >=
366 filter->minimum_silence_buffers)
367 || (filter->minimum_silence_time > 0
368 && filter->consecutive_silence_time >= filter->minimum_silence_time);
369 }
370 if (!filter->silence_detected && consecutive_silence_reached) {
371 if (!filter->silent) {
372 if (GST_BUFFER_PTS_IS_VALID (inbuf)) {
373 GstStructure *s;
374 GstMessage *m;
375 s = gst_structure_new ("removesilence", "silence_detected",
376 G_TYPE_UINT64, GST_BUFFER_PTS (inbuf) - filter->ts_offset, NULL);
377 m = gst_message_new_element (GST_OBJECT (filter), s);
378 gst_element_post_message (GST_ELEMENT (filter), m);
379 }
380 }
381 filter->silence_detected = TRUE;
382 }
383
384 if (filter->remove && consecutive_silence_reached) {
385 GST_DEBUG ("Removing silence");
386 if (filter->squash) {
387 if (GST_BUFFER_DURATION_IS_VALID (inbuf)) {
388 filter->ts_offset += inbuf->duration;
389 } else {
390 GST_WARNING ("Invalid buffer duration: ts_offset not updated");
391 }
392 }
393 return GST_BASE_TRANSFORM_FLOW_DROPPED;
394 }
395
396 } else {
397 filter->consecutive_silence_buffers = 0;
398 filter->consecutive_silence_time = 0;
399 if (filter->silence_detected) {
400 if (!filter->silent) {
401 if (GST_BUFFER_PTS_IS_VALID (inbuf)) {
402 GstStructure *s;
403 GstMessage *m;
404 s = gst_structure_new ("removesilence", "silence_finished",
405 G_TYPE_UINT64, GST_BUFFER_PTS (inbuf) - filter->ts_offset, NULL);
406 m = gst_message_new_element (GST_OBJECT (filter), s);
407 gst_element_post_message (GST_ELEMENT (filter), m);
408 }
409 }
410 filter->silence_detected = FALSE;
411 }
412 }
413
414 if (filter->squash && filter->ts_offset > 0) {
415 if (GST_BUFFER_PTS_IS_VALID (inbuf)) {
416 inbuf = gst_buffer_make_writable (inbuf);
417 GST_BUFFER_PTS (inbuf) -= filter->ts_offset;
418 } else {
419 GST_WARNING ("Invalid buffer pts, update not possibile");
420 }
421 }
422
423 return GST_FLOW_OK;
424 }
425
426 /*Plugin init functions*/
427 static gboolean
plugin_init(GstPlugin * plugin)428 plugin_init (GstPlugin * plugin)
429 {
430 return gst_element_register (plugin, "removesilence", GST_RANK_NONE,
431 gst_remove_silence_get_type ());
432 }
433
434 GST_PLUGIN_DEFINE (GST_VERSION_MAJOR,
435 GST_VERSION_MINOR,
436 removesilence,
437 "Removes silence from an audio stream",
438 plugin_init, VERSION, "LGPL", GST_PACKAGE_NAME, GST_PACKAGE_ORIGIN);
439