[FFmpeg-devel] [PATCH 2/2] lavfi: add audio silencedetect filter.

Nicolas George nicolas.george at normalesup.org
Thu Jan 5 20:29:35 CET 2012


Le sextidi 16 nivôse, an CCXX, Clément Bœsch a écrit :
> ---
>  Changelog                      |    1 +
>  doc/filters.texi               |   23 ++++++
>  libavfilter/Makefile           |    1 +
>  libavfilter/af_silencedetect.c |  161 ++++++++++++++++++++++++++++++++++++++++
>  libavfilter/allfilters.c       |    1 +
>  libavfilter/avfilter.h         |    4 +-
>  6 files changed, 189 insertions(+), 2 deletions(-)
>  create mode 100644 libavfilter/af_silencedetect.c
> 
> diff --git a/Changelog b/Changelog
> index 8b59ac8..80da97b 100644
> --- a/Changelog
> +++ b/Changelog
> @@ -19,6 +19,7 @@ version next:
>  - Avid 1:1 10-bit RGB Packer decoder
>  - v308 Quicktime Uncompressed 4:4:4 encoder and decoder
>  - yuv4 libquicktime packed 4:2:0 encoder and decoder
> +- silencedetect audio filter
>  
>  
>  version 0.9:
> diff --git a/doc/filters.texi b/doc/filters.texi
> index de73e3f..5349fd8 100644
> --- a/doc/filters.texi
> +++ b/doc/filters.texi
> @@ -358,6 +358,29 @@ Note that @command{ffmpeg} integrates a default down-mix (and up-mix) system
>  that should be preferred (see "-ac" option) unless you have very specific
>  needs.
>  

> + at section silencedetect
> +
> +Detect silence in an audio stream.

And do what with it? A few more words may be nice.

> +
> + at table @option
> + at item d, duration
> +Set silence duration until notification (default is 2 seconds).
> +
> + at item noise, n
> +Set noise tolerance. Can be specified in dB or amplitude ratio. Default is
> +-60dB, or 0.001.
> + at end table
> +

> +Detect 5 seconds of silence with -50dB noise tolerance:
> + at example
> +silencedetect=n=-50dB:d=5
> + at end example
> +
> +Detect silence with 0.0001 noise tolerance:
> + at example
> +silencedetect=noise=0.0001
> + at end example

An example with a complete filtergraph may be nice, especially if the use
requires something more complex than just inserting the filter.

> +
>  @section volume
>  
>  Adjust the input audio volume.
> diff --git a/libavfilter/Makefile b/libavfilter/Makefile
> index 0d8f120..bb5748b 100644
> --- a/libavfilter/Makefile
> +++ b/libavfilter/Makefile
> @@ -35,6 +35,7 @@ OBJS-$(CONFIG_ASTREAMSYNC_FILTER)            += af_astreamsync.o
>  OBJS-$(CONFIG_EARWAX_FILTER)                 += af_earwax.o
>  OBJS-$(CONFIG_PAN_FILTER)                    += af_pan.o
>  OBJS-$(CONFIG_VOLUME_FILTER)                 += af_volume.o
> +OBJS-$(CONFIG_SILENCEDETECT_FILTER)          += af_silencedetect.o
>  
>  OBJS-$(CONFIG_ABUFFER_FILTER)                += asrc_abuffer.o
>  OBJS-$(CONFIG_AEVALSRC_FILTER)               += asrc_aevalsrc.o
> diff --git a/libavfilter/af_silencedetect.c b/libavfilter/af_silencedetect.c
> new file mode 100644
> index 0000000..8d68b63
> --- /dev/null
> +++ b/libavfilter/af_silencedetect.c
> @@ -0,0 +1,161 @@

> +/*
> + * This file is part of FFmpeg.
> + *

The copyright line is missing.

> + * 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
> + */
> +
> +/**
> + * @file
> + * Audio silence detector
> + */
> +
> +#include "libavutil/opt.h"
> +#include "avfilter.h"
> +
> +typedef struct {
> +    const AVClass *class;

> +    char *noise_str;

Too bad we can not let the options system parse the option instead of
copying it.

> +    double noise;
> +    int duration;

> +    int nb_null_samples;

int64_t? If the silence if very long, it could overflow.

> +    int silence;
> +} SilenceDetectContext;
> +
> +#define OFFSET(x) offsetof(SilenceDetectContext, x)
> +static const AVOption silencedetect_options[] = {
> +    { "n",         "set noise tolerance",              OFFSET(noise_str), AV_OPT_TYPE_STRING, {.str="-60dB"}, CHAR_MIN, CHAR_MAX },
> +    { "noise",     "set noise tolerance",              OFFSET(noise_str), AV_OPT_TYPE_STRING, {.str="-60dB"}, CHAR_MIN, CHAR_MAX },
> +    { "d",         "set minimum duration in seconds",  OFFSET(duration),  AV_OPT_TYPE_INT,    {.dbl=2},    0, INT_MAX},
> +    { "duration",  "set minimum duration in seconds",  OFFSET(duration),  AV_OPT_TYPE_INT,    {.dbl=2},    0, INT_MAX},
> +    { NULL },
> +};
> +
> +static const char *silencedetect_get_name(void *ctx)
> +{
> +    return "silencedetect";
> +}
> +
> +static const AVClass silencedetect_class = {
> +    .class_name = "SilenceDetectContext",
> +    .item_name  = silencedetect_get_name,
> +    .option     = silencedetect_options,
> +};
> +
> +static av_cold int init(AVFilterContext *ctx, const char *args, void *opaque)
> +{
> +    int ret;
> +    char *tail;
> +    SilenceDetectContext *silence = ctx->priv;
> +
> +    silence->class = &silencedetect_class;
> +    av_opt_set_defaults(silence);
> +
> +    if ((ret = av_set_options_string(silence, args, "=", ":")) < 0) {
> +        av_log(ctx, AV_LOG_ERROR, "Error parsing options string: '%s'\n", args);
> +        return ret;
> +    }
> +
> +    silence->noise = strtod(silence->noise_str, &tail);

> +    if (strcmp(tail, "dB") == 0)

Some people do not like == 0 around here.

> +        silence->noise = pow(10, silence->noise/20);
> +
> +    return 0;
> +}
> +
> +static void filter_samples(AVFilterLink *inlink, AVFilterBufferRef *insamples)
> +{
> +    int i;
> +    SilenceDetectContext *silence = inlink->dst->priv;
> +    const int nb_samples = insamples->audio->nb_samples *
> +        av_get_channel_layout_nb_channels(insamples->audio->channel_layout);
> +
> +    // FIXME: sample rate change will break this

> +    int nbr_samples_notify = inlink->sample_rate * silence->duration;

It seems your code below counts samples per channel, but nbr_samples_notify
is computed without taking the number of channels into account.

Oh, and nbr looks like a French abbreviation, not an English one.

> +
> +    switch (insamples->format) {
> +        case AV_SAMPLE_FMT_DBL: {
> +            double *p = (double *)insamples->data[0];
> +            for (i = 0; i < nb_samples; i++, p++) {
> +                if (*p < silence->noise && *p > -silence->noise) {

> +                    if (!silence->silence) {
> +                        silence->nb_null_samples++;
> +                        if (silence->nb_null_samples == nbr_samples_notify
> +                            && !silence->silence) {

!silence->silence seems duplicated. And possibly even useless: since you
check on silence->nb_null_samples == nbr_samples_notify and not >=, and you
increment each time, it can only be true once.

> +                            av_log(silence, AV_LOG_INFO, "Silence detected around %f sec\n",
> +                                   insamples->pts * av_q2d(inlink->time_base));

I believe insamples->pts can be NOPTS?

> +                            silence->silence = 1;
> +                        }
> +                    }
> +                } else {

> +                    if (silence->silence && silence->nb_null_samples)

Here again, the test could be made simpler: if silence->nb_null_samples >=
nbr_samples_notify.

> +                        av_log(silence, AV_LOG_INFO, "Silence ended (%f sec)\n",
> +                               insamples->pts * av_q2d(inlink->time_base));
> +                    silence->nb_null_samples = silence->silence = 0;
> +                }
> +            }
> +            break;
> +        }
> +    }
> +
> +    avfilter_filter_samples(inlink->dst->outputs[0], insamples);
> +}
> +
> +static int query_formats(AVFilterContext *ctx)
> +{
> +    AVFilterFormats *formats = NULL;
> +    enum AVSampleFormat sample_fmts[] = {
> +        AV_SAMPLE_FMT_DBL,
> +        AV_SAMPLE_FMT_NONE
> +    };
> +    int packing_fmts[] = { AVFILTER_PACKED, -1 };
> +
> +    formats = avfilter_make_all_channel_layouts();
> +    if (!formats)
> +        return AVERROR(ENOMEM);
> +    avfilter_set_common_channel_layouts(ctx, formats);
> +
> +    formats = avfilter_make_format_list(sample_fmts);
> +    if (!formats)
> +        return AVERROR(ENOMEM);
> +    avfilter_set_common_sample_formats(ctx, formats);
> +
> +    formats = avfilter_make_format_list(packing_fmts);
> +    if (!formats)
> +        return AVERROR(ENOMEM);
> +    avfilter_set_common_packing_formats(ctx, formats);
> +
> +    return 0;
> +}
> +
> +AVFilter avfilter_af_silencedetect = {
> +    .name          = "silencedetect",
> +    .description   = NULL_IF_CONFIG_SMALL("Detect silence."),
> +    .priv_size     = sizeof(SilenceDetectContext),
> +    .init          = init,
> +    .query_formats = query_formats,
> +
> +    .inputs = (const AVFilterPad[]) {
> +        { .name             = "default",
> +          .type             = AVMEDIA_TYPE_AUDIO,
> +          .get_audio_buffer = avfilter_null_get_audio_buffer,
> +          .filter_samples   = filter_samples, },

> +        { .name = NULL}

Is it on purpose that most filters lack the space between the NULL and the
closing brace?

> +    },
> +    .outputs = (const AVFilterPad[]) {
> +        { .name = "default",
> +          .type = AVMEDIA_TYPE_AUDIO, },
> +        { .name = NULL}
> +    },

Did you consider making it a sink rather than a filter?

> +};
> diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
> index 621568e..a863a93 100644
> --- a/libavfilter/allfilters.c
> +++ b/libavfilter/allfilters.c
> @@ -45,6 +45,7 @@ void avfilter_register_all(void)
>      REGISTER_FILTER (EARWAX,      earwax,      af);
>      REGISTER_FILTER (PAN,         pan,         af);
>      REGISTER_FILTER (VOLUME,      volume,      af);
> +    REGISTER_FILTER (SILENCEDETECT, silencedetect, af);
>  
>      REGISTER_FILTER (ABUFFER,     abuffer,     asrc);
>      REGISTER_FILTER (AEVALSRC,    aevalsrc,    asrc);
> diff --git a/libavfilter/avfilter.h b/libavfilter/avfilter.h
> index d88d3ab..9c79597 100644
> --- a/libavfilter/avfilter.h
> +++ b/libavfilter/avfilter.h
> @@ -30,8 +30,8 @@
>  #include "libavcodec/avcodec.h"
>  
>  #define LIBAVFILTER_VERSION_MAJOR  2
> -#define LIBAVFILTER_VERSION_MINOR 57
> -#define LIBAVFILTER_VERSION_MICRO 101
> +#define LIBAVFILTER_VERSION_MINOR 58
> +#define LIBAVFILTER_VERSION_MICRO 100
>  
>  #define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \
>                                                 LIBAVFILTER_VERSION_MINOR, \

Regards,

-- 
  Nicolas George
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20120105/2f3afd46/attachment.asc>


More information about the ffmpeg-devel mailing list