Subversion Repositories Kolibri OS

Rev

Blame | Last modification | View Log | RSS feed

  1. /*
  2.  * Copyright (c) 2012 Nicolas George
  3.  *
  4.  * This file is part of FFmpeg.
  5.  *
  6.  * FFmpeg is free software; you can redistribute it and/or
  7.  * modify it under the terms of the GNU Lesser General Public License
  8.  * as published by the Free Software Foundation; either
  9.  * version 2.1 of the License, or (at your option) any later version.
  10.  *
  11.  * FFmpeg 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
  14.  * GNU Lesser General Public License for more details.
  15.  *
  16.  * You should have received a copy of the GNU Lesser General Public License
  17.  * along with FFmpeg; if not, write to the Free Software Foundation, Inc.,
  18.  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19.  */
  20.  
  21. #include "libavutil/channel_layout.h"
  22. #include "libavutil/avassert.h"
  23. #include "audio.h"
  24. #include "avfilter.h"
  25. #include "internal.h"
  26.  
  27. typedef struct {
  28.     /**
  29.      * Number of samples at each PCM value.
  30.      * histogram[0x8000 + i] is the number of samples at value i.
  31.      * The extra element is there for symmetry.
  32.      */
  33.     uint64_t histogram[0x10001];
  34. } VolDetectContext;
  35.  
  36. static int query_formats(AVFilterContext *ctx)
  37. {
  38.     static const enum AVSampleFormat sample_fmts[] = {
  39.         AV_SAMPLE_FMT_S16,
  40.         AV_SAMPLE_FMT_S16P,
  41.         AV_SAMPLE_FMT_NONE
  42.     };
  43.     AVFilterFormats *formats;
  44.  
  45.     if (!(formats = ff_make_format_list(sample_fmts)))
  46.         return AVERROR(ENOMEM);
  47.     return ff_set_common_formats(ctx, formats);
  48. }
  49.  
  50. static int filter_frame(AVFilterLink *inlink, AVFrame *samples)
  51. {
  52.     AVFilterContext *ctx = inlink->dst;
  53.     VolDetectContext *vd = ctx->priv;
  54.     int64_t layout  = samples->channel_layout;
  55.     int nb_samples  = samples->nb_samples;
  56.     int nb_channels = av_get_channel_layout_nb_channels(layout);
  57.     int nb_planes   = nb_channels;
  58.     int plane, i;
  59.     int16_t *pcm;
  60.  
  61.     if (!av_sample_fmt_is_planar(samples->format)) {
  62.         nb_samples *= nb_channels;
  63.         nb_planes = 1;
  64.     }
  65.     for (plane = 0; plane < nb_planes; plane++) {
  66.         pcm = (int16_t *)samples->extended_data[plane];
  67.         for (i = 0; i < nb_samples; i++)
  68.             vd->histogram[pcm[i] + 0x8000]++;
  69.     }
  70.  
  71.     return ff_filter_frame(inlink->dst->outputs[0], samples);
  72. }
  73.  
  74. #define MAX_DB 91
  75.  
  76. static inline double logdb(uint64_t v)
  77. {
  78.     double d = v / (double)(0x8000 * 0x8000);
  79.     if (!v)
  80.         return MAX_DB;
  81.     return log(d) * -4.3429448190325182765112891891660508229; /* -10/log(10) */
  82. }
  83.  
  84. static void print_stats(AVFilterContext *ctx)
  85. {
  86.     VolDetectContext *vd = ctx->priv;
  87.     int i, max_volume, shift;
  88.     uint64_t nb_samples = 0, power = 0, nb_samples_shift = 0, sum = 0;
  89.     uint64_t histdb[MAX_DB + 1] = { 0 };
  90.  
  91.     for (i = 0; i < 0x10000; i++)
  92.         nb_samples += vd->histogram[i];
  93.     av_log(ctx, AV_LOG_INFO, "n_samples: %"PRId64"\n", nb_samples);
  94.     if (!nb_samples)
  95.         return;
  96.  
  97.     /* If nb_samples > 1<<34, there is a risk of overflow in the
  98.        multiplication or the sum: shift all histogram values to avoid that.
  99.        The total number of samples must be recomputed to avoid rounding
  100.        errors. */
  101.     shift = av_log2(nb_samples >> 33);
  102.     for (i = 0; i < 0x10000; i++) {
  103.         nb_samples_shift += vd->histogram[i] >> shift;
  104.         power += (i - 0x8000) * (i - 0x8000) * (vd->histogram[i] >> shift);
  105.     }
  106.     if (!nb_samples_shift)
  107.         return;
  108.     power = (power + nb_samples_shift / 2) / nb_samples_shift;
  109.     av_assert0(power <= 0x8000 * 0x8000);
  110.     av_log(ctx, AV_LOG_INFO, "mean_volume: %.1f dB\n", -logdb(power));
  111.  
  112.     max_volume = 0x8000;
  113.     while (max_volume > 0 && !vd->histogram[0x8000 + max_volume] &&
  114.                              !vd->histogram[0x8000 - max_volume])
  115.         max_volume--;
  116.     av_log(ctx, AV_LOG_INFO, "max_volume: %.1f dB\n", -logdb(max_volume * max_volume));
  117.  
  118.     for (i = 0; i < 0x10000; i++)
  119.         histdb[(int)logdb((i - 0x8000) * (i - 0x8000))] += vd->histogram[i];
  120.     for (i = 0; i <= MAX_DB && !histdb[i]; i++);
  121.     for (; i <= MAX_DB && sum < nb_samples / 1000; i++) {
  122.         av_log(ctx, AV_LOG_INFO, "histogram_%ddb: %"PRId64"\n", i, histdb[i]);
  123.         sum += histdb[i];
  124.     }
  125. }
  126.  
  127. static av_cold void uninit(AVFilterContext *ctx)
  128. {
  129.     print_stats(ctx);
  130. }
  131.  
  132. static const AVFilterPad volumedetect_inputs[] = {
  133.     {
  134.         .name         = "default",
  135.         .type         = AVMEDIA_TYPE_AUDIO,
  136.         .filter_frame = filter_frame,
  137.     },
  138.     { NULL }
  139. };
  140.  
  141. static const AVFilterPad volumedetect_outputs[] = {
  142.     {
  143.         .name = "default",
  144.         .type = AVMEDIA_TYPE_AUDIO,
  145.     },
  146.     { NULL }
  147. };
  148.  
  149. AVFilter ff_af_volumedetect = {
  150.     .name          = "volumedetect",
  151.     .description   = NULL_IF_CONFIG_SMALL("Detect audio volume."),
  152.     .priv_size     = sizeof(VolDetectContext),
  153.     .query_formats = query_formats,
  154.     .uninit        = uninit,
  155.     .inputs        = volumedetect_inputs,
  156.     .outputs       = volumedetect_outputs,
  157. };
  158.