add silenceremove filter

Signed-off-by: Paul B Mahol <onemda@gmail.com>

8 files changed, 558 insertions, 2 deletions

diff --git a/Changelog b/Changelog
index 72c783aee9..c5185f7eec 100644
--- a/Changelog
+++ b/Changelog
@@ -13,6 +13,7 @@ version <next>:
 - added codecview filter to visualize information exported by some codecs
 - Matroska 3D support thorugh side data
 - HTML generation using texi2html is deprecated in favor of makeinfo/texi2any
+- silenceremove filter
 
 
 version 2.3:
diff --git a/MAINTAINERS b/MAINTAINERS
index eb1f98bdd9..8af02ae62a 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -343,6 +343,7 @@ Filters:
   af_compand.c                          Paul B Mahol
   af_ladspa.c                           Paul B Mahol
   af_pan.c                              Nicolas George
+  af_silenceremove.c                    Paul B Mahol
   avf_avectorscope.c                    Paul B Mahol
   avf_showcqt.c                         Muhammad Faiz
   vf_blend.c                            Paul B Mahol
diff --git a/RELEASE_NOTES b/RELEASE_NOTES
index 15eb81b476..113cc5e97d 100644
--- a/RELEASE_NOTES
+++ b/RELEASE_NOTES
@@ -42,6 +42,7 @@
     • ported lenscorrection filter from frei0r filter
     • large optimizations in dctdnoiz to make it usable
     • added codecview filter to visualize information exported by some codecs
+    • added silenceremove filter
 
    ┌────────────────────────────┐
    │ libavutil                  │
diff --git a/doc/filters.texi b/doc/filters.texi
index d13278ef71..627f112724 100644
--- a/doc/filters.texi
+++ b/doc/filters.texi
@@ -1875,6 +1875,75 @@ ffmpeg -i silence.mp3 -af silencedetect=noise=0.0001 -f null -
 @end example
 @end itemize
 
+@section silenceremove
+
+Remove silence from the beginning, middle or end of the audio.
+
+The filter accepts the following options:
+
+@table @option
+@item start_periods
+This value is used to indicate if audio should be trimmed at beginning of
+the audio. A value of zero indicates no silence should be trimmed from the
+beginning. When specifying a non-zero value, it trims audio up until it
+finds non-silence. Normally, when trimming silence from beginning of audio
+the @var{start_periods} will be @code{1} but it can be increased to higher
+values to trim all audio up to specific count of non-silence periods.
+Default value is @code{0}.
+
+@item start_duration
+Specify the amount of time that non-silence must be detected before it stops
+trimming audio. By increasing the duration, bursts of noises can be treated
+as silence and trimmed off. Default value is @code{0}.
+
+@item start_threshold
+This indicates what sample value should be treated as silence. For digital
+audio, a value of @code{0} may be fine but for audio recorded from analog,
+you may wish to increase the value to account for background noise.
+Can be specified in dB (in case "dB" is appended to the specified value)
+or amplitude ratio. Default value is @code{0}.
+
+@item stop_periods
+Set the count for trimming silence from the end of audio.
+To remove silence from the middle of a file, specify a @var{stop_periods}
+that is negative. This value is then threated as a positive value and is
+used to indicate the effect should restart processing as specified by
+@var{start_periods}, making it suitable for removing periods of silence
+in the middle of the audio.
+Default value is @code{0}.
+
+@item stop_duration
+Specify a duration of silence that must exist before audio is not copied any
+more. By specifying a higher duration, silence that is wanted can be left in
+the audio.
+Default value is @code{0}.
+
+@item stop_threshold
+This is the same as @option{start_threshold} but for trimming silence from
+the end of audio.
+Can be specified in dB (in case "dB" is appended to the specified value)
+or amplitude ratio. Default value is @code{0}.
+
+@item leave_silence
+This indicate that @var{stop_duration} length of audio should be left intact
+at the beginning of each period of silence.
+For example, if you want to remove long pauses between words but do not want
+to remove the pauses completely. Default value is @code{0}.
+
+@end table
+
+@subsection Examples
+
+@itemize
+@item
+The following example shows how this filter can be used to start a recording
+that does not contain the delay at the start which usually occurs between
+pressing the record button and the start of the performance:
+@example
+silenceremove=1:5:0.02
+@end example
+@end itemize
+
 @section treble
 
 Boost or cut treble (upper) frequencies of the audio using a two-pole
diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index ce71ce1a24..3241b76f10 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -78,6 +78,7 @@ OBJS-$(CONFIG_PAN_FILTER)                    += af_pan.o
 OBJS-$(CONFIG_REPLAYGAIN_FILTER)             += af_replaygain.o
 OBJS-$(CONFIG_RESAMPLE_FILTER)               += af_resample.o
 OBJS-$(CONFIG_SILENCEDETECT_FILTER)          += af_silencedetect.o
+OBJS-$(CONFIG_SILENCEREMOVE_FILTER)          += af_silenceremove.o
 OBJS-$(CONFIG_TREBLE_FILTER)                 += af_biquads.o
 OBJS-$(CONFIG_VOLUME_FILTER)                 += af_volume.o
 OBJS-$(CONFIG_VOLUMEDETECT_FILTER)           += af_volumedetect.o
diff --git a/libavfilter/af_silenceremove.c b/libavfilter/af_silenceremove.c
new file mode 100644
index 0000000000..3f6cb7a550
--- /dev/null
+++ b/libavfilter/af_silenceremove.c
@@ -0,0 +1,482 @@
+/*
+ * Copyright (c) 2001 Heikki Leinonen
+ * Copyright (c) 2001 Chris Bagwell
+ * Copyright (c) 2003 Donnie Smith
+ * Copyright (c) 2014 Paul B Mahol
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <float.h> /* DBL_MAX */
+
+#include "libavutil/opt.h"
+#include "libavutil/timestamp.h"
+#include "audio.h"
+#include "formats.h"
+#include "avfilter.h"
+#include "internal.h"
+
+enum SilenceMode {
+    SILENCE_TRIM,
+    SILENCE_TRIM_FLUSH,
+    SILENCE_COPY,
+    SILENCE_COPY_FLUSH,
+    SILENCE_STOP
+};
+
+typedef struct SilenceRemoveContext {
+    const AVClass *class;
+
+    enum SilenceMode mode;
+
+    int start_periods;
+    int64_t start_duration;
+    double start_threshold;
+
+    int stop_periods;
+    int64_t stop_duration;
+    double stop_threshold;
+
+    double *start_holdoff;
+    size_t start_holdoff_offset;
+    size_t start_holdoff_end;
+    int    start_found_periods;
+
+    double *stop_holdoff;
+    size_t stop_holdoff_offset;
+    size_t stop_holdoff_end;
+    int    stop_found_periods;
+
+    double *window;
+    double *window_current;
+    double *window_end;
+    int window_size;
+    double rms_sum;
+
+    int leave_silence;
+    int restart;
+    int64_t next_pts;
+} SilenceRemoveContext;
+
+#define OFFSET(x) offsetof(SilenceRemoveContext, x)
+#define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_AUDIO_PARAM
+static const AVOption silenceremove_options[] = {
+    { "start_periods",   NULL, OFFSET(start_periods),   AV_OPT_TYPE_INT,      {.i64=0},     0,    9000, FLAGS },
+    { "start_duration",  NULL, OFFSET(start_duration),  AV_OPT_TYPE_DURATION, {.i64=0},     0,    9000, FLAGS },
+    { "start_threshold", NULL, OFFSET(start_threshold), AV_OPT_TYPE_DOUBLE,   {.dbl=0},     0, DBL_MAX, FLAGS },
+    { "stop_periods",    NULL, OFFSET(stop_periods),    AV_OPT_TYPE_INT,      {.i64=0}, -9000,    9000, FLAGS },
+    { "stop_duration",   NULL, OFFSET(stop_duration),   AV_OPT_TYPE_DURATION, {.i64=0},     0,    9000, FLAGS },
+    { "stop_threshold",  NULL, OFFSET(stop_threshold),  AV_OPT_TYPE_DOUBLE,   {.dbl=0},     0, DBL_MAX, FLAGS },
+    { "leave_silence",   NULL, OFFSET(leave_silence),   AV_OPT_TYPE_INT,      {.i64=0},     0,       1, FLAGS },
+    { NULL }
+};
+
+AVFILTER_DEFINE_CLASS(silenceremove);
+
+static av_cold int init(AVFilterContext *ctx)
+{
+    SilenceRemoveContext *s = ctx->priv;
+
+    if (s->stop_periods < 0) {
+        s->stop_periods = -s->stop_periods;
+        s->restart = 1;
+    }
+
+    return 0;
+}
+
+static void clear_rms(SilenceRemoveContext *s)
+{
+    memset(s->window, 0, s->window_size * sizeof(*s->window));
+
+    s->window_current = s->window;
+    s->window_end = s->window + s->window_size;
+    s->rms_sum = 0;
+}
+
+static int config_input(AVFilterLink *inlink)
+{
+    AVFilterContext *ctx = inlink->dst;
+    SilenceRemoveContext *s = ctx->priv;
+
+    s->window_size = (inlink->sample_rate / 50) * inlink->channels;
+    s->window = av_malloc_array(s->window_size, sizeof(*s->window));
+    if (!s->window)
+        return AVERROR(ENOMEM);
+
+    clear_rms(s);
+
+    s->start_duration = av_rescale(s->start_duration, inlink->sample_rate,
+                                   AV_TIME_BASE);
+    s->stop_duration  = av_rescale(s->stop_duration, inlink->sample_rate,
+                                   AV_TIME_BASE);
+
+    s->start_holdoff = av_malloc_array(FFMAX(s->start_duration, 1),
+                                       sizeof(*s->start_holdoff) *
+                                       inlink->channels);
+    if (!s->start_holdoff)
+        return AVERROR(ENOMEM);
+
+    s->start_holdoff_offset = 0;
+    s->start_holdoff_end    = 0;
+    s->start_found_periods  = 0;
+
+    s->stop_holdoff = av_malloc_array(FFMAX(s->stop_duration, 1),
+                                      sizeof(*s->stop_holdoff) *
+                                      inlink->channels);
+    if (!s->stop_holdoff)
+        return AVERROR(ENOMEM);
+
+    s->stop_holdoff_offset = 0;
+    s->stop_holdoff_end    = 0;
+    s->stop_found_periods  = 0;
+
+    if (s->start_periods)
+        s->mode = SILENCE_TRIM;
+    else
+        s->mode = SILENCE_COPY;
+
+    return 0;
+}
+
+static int config_output(AVFilterLink *outlink)
+{
+    outlink->flags |= FF_LINK_FLAG_REQUEST_LOOP;
+
+    return 0;
+}
+
+static double compute_rms(SilenceRemoveContext *s, double sample)
+{
+    double new_sum;
+
+    new_sum  = s->rms_sum;
+    new_sum -= *s->window_current;
+    new_sum += sample * sample;
+
+    return sqrt(new_sum / s->window_size);
+}
+
+static void update_rms(SilenceRemoveContext *s, double sample)
+{
+    s->rms_sum -= *s->window_current;
+    *s->window_current = sample * sample;
+    s->rms_sum += *s->window_current;
+
+    s->window_current++;
+    if (s->window_current >= s->window_end)
+        s->window_current = s->window;
+}
+
+static void flush(AVFrame *out, AVFilterLink *outlink,
+                  int *nb_samples_written, int *ret)
+{
+    if (*nb_samples_written) {
+        out->nb_samples = *nb_samples_written / outlink->channels;
+        *ret = ff_filter_frame(outlink, out);
+        *nb_samples_written = 0;
+    } else {
+        av_frame_free(&out);
+    }
+}
+
+static int filter_frame(AVFilterLink *inlink, AVFrame *in)
+{
+    AVFilterContext *ctx = inlink->dst;
+    AVFilterLink *outlink = ctx->outputs[0];
+    SilenceRemoveContext *s = ctx->priv;
+    int i, j, threshold, ret = 0;
+    int nbs, nb_samples_read, nb_samples_written;
+    double *obuf, *ibuf = (double *)in->data[0];
+    AVFrame *out;
+
+    nb_samples_read = nb_samples_written = 0;
+
+    switch (s->mode) {
+    case SILENCE_TRIM:
+silence_trim:
+        nbs = in->nb_samples - nb_samples_read / inlink->channels;
+        if (!nbs)
+            break;
+
+        for (i = 0; i < nbs; i++) {
+            threshold = 0;
+            for (j = 0; j < inlink->channels; j++) {
+                threshold |= compute_rms(s, ibuf[j]) > s->start_threshold;
+            }
+
+            if (threshold) {
+                for (j = 0; j < inlink->channels; j++) {
+                    update_rms(s, *ibuf);
+                    s->start_holdoff[s->start_holdoff_end++] = *ibuf++;
+                    nb_samples_read++;
+                }
+
+                if (s->start_holdoff_end >= s->start_duration * inlink->channels) {
+                    if (++s->start_found_periods >= s->start_periods) {
+                        s->mode = SILENCE_TRIM_FLUSH;
+                        goto silence_trim_flush;
+                    }
+
+                    s->start_holdoff_offset = 0;
+                    s->start_holdoff_end = 0;
+                }
+            } else {
+                s->start_holdoff_end = 0;
+
+                for (j = 0; j < inlink->channels; j++)
+                    update_rms(s, ibuf[j]);
+
+                ibuf += inlink->channels;
+                nb_samples_read += inlink->channels;
+            }
+        }
+        break;
+
+    case SILENCE_TRIM_FLUSH:
+silence_trim_flush:
+        nbs  = s->start_holdoff_end - s->start_holdoff_offset;
+        nbs -= nbs % inlink->channels;
+        if (!nbs)
+            break;
+
+        out = ff_get_audio_buffer(inlink, nbs / inlink->channels);
+        if (!out) {
+            av_frame_free(&in);
+            return AVERROR(ENOMEM);
+        }
+
+        memcpy(out->data[0], &s->start_holdoff[s->start_holdoff_offset],
+               nbs * sizeof(double));
+        s->start_holdoff_offset += nbs;
+
+        ret = ff_filter_frame(outlink, out);
+
+        if (s->start_holdoff_offset == s->start_holdoff_end) {
+            s->start_holdoff_offset = 0;
+            s->start_holdoff_end = 0;
+            s->mode = SILENCE_COPY;
+            goto silence_copy;
+        }
+        break;
+
+    case SILENCE_COPY:
+silence_copy:
+        nbs = in->nb_samples - nb_samples_read / inlink->channels;
+        if (!nbs)
+            break;
+
+        out = ff_get_audio_buffer(inlink, nbs);
+        if (!out) {
+            av_frame_free(&in);
+            return AVERROR(ENOMEM);
+        }
+        obuf = (double *)out->data[0];
+
+        if (s->stop_periods) {
+            for (i = 0; i < nbs; i++) {
+                threshold = 1;
+                for (j = 0; j < inlink->channels; j++)
+                    threshold &= compute_rms(s, ibuf[j]) > s->stop_threshold;
+
+                if (threshold && s->stop_holdoff_end && !s->leave_silence) {
+                    s->mode = SILENCE_COPY_FLUSH;
+                    flush(out, outlink, &nb_samples_written, &ret);
+                    goto silence_copy_flush;
+                } else if (threshold) {
+                    for (j = 0; j < inlink->channels; j++) {
+                        update_rms(s, *ibuf);
+                        *obuf++ = *ibuf++;
+                        nb_samples_read++;
+                        nb_samples_written++;
+                    }
+                } else if (!threshold) {
+                    for (j = 0; j < inlink->channels; j++) {
+                        update_rms(s, *ibuf);
+                        if (s->leave_silence) {
+                            *obuf++ = *ibuf;
+                            nb_samples_written++;
+                        }
+
+                        s->stop_holdoff[s->stop_holdoff_end++] = *ibuf++;
+                        nb_samples_read++;
+                    }
+
+                    if (s->stop_holdoff_end >= s->stop_duration * inlink->channels) {
+                        if (++s->stop_found_periods >= s->stop_periods) {
+                            s->stop_holdoff_offset = 0;
+                            s->stop_holdoff_end = 0;
+
+                            if (!s->restart) {
+                                s->mode = SILENCE_STOP;
+                                flush(out, outlink, &nb_samples_written, &ret);
+                                goto silence_stop;
+                            } else {
+                                s->stop_found_periods = 0;
+                                s->start_found_periods = 0;
+                                s->start_holdoff_offset = 0;
+                                s->start_holdoff_end = 0;
+                                clear_rms(s);
+                                s->mode = SILENCE_TRIM;
+                                flush(out, outlink, &nb_samples_written, &ret);
+                                goto silence_trim;
+                            }
+                        } else {
+                            s->mode = SILENCE_COPY_FLUSH;
+                            flush(out, outlink, &nb_samples_written, &ret);
+                            goto silence_copy_flush;
+                        }
+                        flush(out, outlink, &nb_samples_written, &ret);
+                        break;
+                    }
+                }
+            }
+            flush(out, outlink, &nb_samples_written, &ret);
+        } else {
+            memcpy(obuf, ibuf, sizeof(double) * nbs * inlink->channels);
+            ret = ff_filter_frame(outlink, out);
+        }
+        break;
+
+    case SILENCE_COPY_FLUSH:
+silence_copy_flush:
+        nbs  = s->stop_holdoff_end - s->stop_holdoff_offset;
+        nbs -= nbs % inlink->channels;
+        if (!nbs)
+            break;
+
+        out = ff_get_audio_buffer(inlink, nbs / inlink->channels);
+        if (!out) {
+            av_frame_free(&in);
+            return AVERROR(ENOMEM);
+        }
+
+        memcpy(out->data[0], &s->stop_holdoff[s->stop_holdoff_offset],
+               nbs * sizeof(double));
+        s->stop_holdoff_offset += nbs;
+
+        ret = ff_filter_frame(outlink, out);
+
+        if (s->stop_holdoff_offset == s->stop_holdoff_end) {
+            s->stop_holdoff_offset = 0;
+            s->stop_holdoff_end = 0;
+            s->mode = SILENCE_COPY;
+            goto silence_copy;
+        }
+        break;
+    case SILENCE_STOP:
+silence_stop:
+        break;
+    }
+
+    av_frame_free(&in);
+
+    return ret;
+}
+
+static int request_frame(AVFilterLink *outlink)
+{
+    AVFilterContext *ctx = outlink->src;
+    SilenceRemoveContext *s = ctx->priv;
+    int ret;
+
+    ret = ff_request_frame(ctx->inputs[0]);
+    if (ret == AVERROR_EOF && (s->mode == SILENCE_COPY_FLUSH ||
+                               s->mode == SILENCE_COPY)) {
+        int nbs = s->stop_holdoff_end - s->stop_holdoff_offset;
+        if (nbs) {
+            AVFrame *frame;
+
+            frame = ff_get_audio_buffer(outlink, nbs / outlink->channels);
+            if (!frame)
+                return AVERROR(ENOMEM);
+
+            memcpy(frame->data[0], &s->stop_holdoff[s->stop_holdoff_offset],
+                   nbs * sizeof(double));
+            ret = ff_filter_frame(ctx->inputs[0], frame);
+        }
+        s->mode = SILENCE_STOP;
+    }
+    return ret;
+}
+
+static int query_formats(AVFilterContext *ctx)
+{
+    AVFilterFormats *formats = NULL;
+    AVFilterChannelLayouts *layouts = NULL;
+    static const enum AVSampleFormat sample_fmts[] = {
+        AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_NONE
+    };
+
+    layouts = ff_all_channel_layouts();
+    if (!layouts)
+        return AVERROR(ENOMEM);
+    ff_set_common_channel_layouts(ctx, layouts);
+
+    formats = ff_make_format_list(sample_fmts);
+    if (!formats)
+        return AVERROR(ENOMEM);
+    ff_set_common_formats(ctx, formats);
+
+    formats = ff_all_samplerates();
+    if (!formats)
+        return AVERROR(ENOMEM);
+    ff_set_common_samplerates(ctx, formats);
+
+    return 0;
+}
+
+static av_cold void uninit(AVFilterContext *ctx)
+{
+    SilenceRemoveContext *s = ctx->priv;
+
+    av_freep(&s->start_holdoff);
+    av_freep(&s->stop_holdoff);
+    av_freep(&s->window);
+}
+
+static const AVFilterPad silenceremove_inputs[] = {
+    {
+        .name         = "default",
+        .type         = AVMEDIA_TYPE_AUDIO,
+        .config_props = config_input,
+        .filter_frame = filter_frame,
+    },
+    { NULL }
+};
+
+static const AVFilterPad silenceremove_outputs[] = {
+    {
+        .name          = "default",
+        .type          = AVMEDIA_TYPE_AUDIO,
+        .config_props  = config_output,
+        .request_frame = request_frame,
+    },
+    { NULL }
+};
+
+AVFilter ff_af_silenceremove = {
+    .name          = "silenceremove",
+    .description   = NULL_IF_CONFIG_SMALL("Remove silence."),
+    .priv_size     = sizeof(SilenceRemoveContext),
+    .priv_class    = &silenceremove_class,
+    .init          = init,
+    .uninit        = uninit,
+    .query_formats = query_formats,
+    .inputs        = silenceremove_inputs,
+    .outputs       = silenceremove_outputs,
+};
diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
index 8fe2020901..670f2d1c77 100644
--- a/libavfilter/allfilters.c
+++ b/libavfilter/allfilters.c
@@ -96,6 +96,7 @@ void avfilter_register_all(void)
     REGISTER_FILTER(REPLAYGAIN,     replaygain,     af);
     REGISTER_FILTER(RESAMPLE,       resample,       af);
     REGISTER_FILTER(SILENCEDETECT,  silencedetect,  af);
+    REGISTER_FILTER(SILENCEREMOVE,  silenceremove,  af);
     REGISTER_FILTER(TREBLE,         treble,         af);
     REGISTER_FILTER(VOLUME,         volume,         af);
     REGISTER_FILTER(VOLUMEDETECT,   volumedetect,   af);
diff --git a/libavfilter/version.h b/libavfilter/version.h
index 7c9883d494..6ebb1ce609 100644
--- a/libavfilter/version.h
+++ b/libavfilter/version.h
@@ -30,8 +30,8 @@
 #include "libavutil/version.h"
 
 #define LIBAVFILTER_VERSION_MAJOR  5
-#define LIBAVFILTER_VERSION_MINOR  0
-#define LIBAVFILTER_VERSION_MICRO 103
+#define LIBAVFILTER_VERSION_MINOR  1
+#define LIBAVFILTER_VERSION_MICRO 100
 
 #define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \
                                                LIBAVFILTER_VERSION_MINOR, \