2 * Copyright (c) 2009 Rob Sykes <robs@users.sourceforge.net>
3 * Copyright (c) 2013 Paul B Mahol
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 #include "libavutil/opt.h"
29 typedef struct ChannelStats {
31 double sigma_x, sigma_x2;
32 double avg_sigma_x2, min_sigma_x2, max_sigma_x2;
34 double min_run, max_run;
35 double min_runs, max_runs;
36 uint64_t min_count, max_count;
42 ChannelStats *chstats;
49 #define OFFSET(x) offsetof(AudioStatsContext, x)
50 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
52 static const AVOption astats_options[] = {
53 { "length", "set the window length", OFFSET(time_constant), AV_OPT_TYPE_DOUBLE, {.dbl=.05}, .01, 10, FLAGS },
57 AVFILTER_DEFINE_CLASS(astats);
59 static int query_formats(AVFilterContext *ctx)
61 AVFilterFormats *formats;
62 AVFilterChannelLayouts *layouts;
63 static const enum AVSampleFormat sample_fmts[] = {
64 AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_DBLP,
68 layouts = ff_all_channel_layouts();
70 return AVERROR(ENOMEM);
71 ff_set_common_channel_layouts(ctx, layouts);
73 formats = ff_make_format_list(sample_fmts);
75 return AVERROR(ENOMEM);
76 ff_set_common_formats(ctx, formats);
78 formats = ff_all_samplerates();
80 return AVERROR(ENOMEM);
81 ff_set_common_samplerates(ctx, formats);
86 static int config_output(AVFilterLink *outlink)
88 AudioStatsContext *s = outlink->src->priv;
91 s->chstats = av_calloc(sizeof(*s->chstats), outlink->channels);
93 return AVERROR(ENOMEM);
94 s->nb_channels = outlink->channels;
95 s->mult = exp((-1 / s->time_constant / outlink->sample_rate));
96 s->tc_samples = 5 * s->time_constant * outlink->sample_rate + .5;
98 for (c = 0; c < s->nb_channels; c++) {
99 ChannelStats *p = &s->chstats[c];
101 p->min = p->min_sigma_x2 = DBL_MAX;
102 p->max = p->max_sigma_x2 = DBL_MIN;
108 static inline void stat(AudioStatsContext *s, ChannelStats *p, double d)
115 } else if (d == p->min) {
117 p->min_run = d == p->last ? p->min_run + 1 : 1;
118 } else if (p->last == p->min) {
119 p->min_runs += p->min_run * p->min_run;
127 } else if (d == p->max) {
129 p->max_run = d == p->last ? p->max_run + 1 : 1;
130 } else if (p->last == p->max) {
131 p->max_runs += p->max_run * p->max_run;
135 p->sigma_x2 += d * d;
136 p->avg_sigma_x2 = p->avg_sigma_x2 * s->mult + (1.0 - s->mult) * d * d;
139 if (p->nb_samples >= s->tc_samples) {
140 p->max_sigma_x2 = FFMAX(p->max_sigma_x2, p->avg_sigma_x2);
141 p->min_sigma_x2 = FFMIN(p->min_sigma_x2, p->avg_sigma_x2);
146 static int filter_frame(AVFilterLink *inlink, AVFrame *buf)
148 AudioStatsContext *s = inlink->dst->priv;
149 const int channels = s->nb_channels;
153 switch (inlink->format) {
154 case AV_SAMPLE_FMT_DBLP:
155 for (c = 0; c < channels; c++) {
156 ChannelStats *p = &s->chstats[c];
157 src = (const double *)buf->extended_data[c];
159 for (i = 0; i < buf->nb_samples; i++, src++)
163 case AV_SAMPLE_FMT_DBL:
164 src = (const double *)buf->extended_data[0];
166 for (i = 0; i < buf->nb_samples; i++) {
167 for (c = 0; c < channels; c++, src++)
168 stat(s, &s->chstats[c], *src);
173 return ff_filter_frame(inlink->dst->outputs[0], buf);
176 #define LINEAR_TO_DB(x) (log10(x) * 20)
178 static void print_stats(AVFilterContext *ctx)
180 AudioStatsContext *s = ctx->priv;
181 uint64_t min_count = 0, max_count = 0, nb_samples = 0;
182 double min_runs = 0, max_runs = 0,
183 min = DBL_MAX, max = DBL_MIN,
187 min_sigma_x2 = DBL_MAX,
188 max_sigma_x2 = DBL_MIN;
191 for (c = 0; c < s->nb_channels; c++) {
192 ChannelStats *p = &s->chstats[c];
194 if (p->nb_samples < s->tc_samples)
195 p->min_sigma_x2 = p->max_sigma_x2 = p->sigma_x2 / p->nb_samples;
197 min = FFMIN(min, p->min);
198 max = FFMAX(max, p->max);
199 min_sigma_x2 = FFMIN(min_sigma_x2, p->min_sigma_x2);
200 max_sigma_x2 = FFMAX(max_sigma_x2, p->max_sigma_x2);
201 sigma_x += p->sigma_x;
202 sigma_x2 += p->sigma_x2;
203 min_count += p->min_count;
204 max_count += p->max_count;
205 min_runs += p->min_runs;
206 max_runs += p->max_runs;
207 nb_samples += p->nb_samples;
208 if (fabs(p->sigma_x) > fabs(max_sigma_x))
209 max_sigma_x = p->sigma_x;
211 av_log(ctx, AV_LOG_INFO, "Channel: %d\n", c + 1);
212 av_log(ctx, AV_LOG_INFO, "DC offset: %f\n", p->sigma_x / p->nb_samples);
213 av_log(ctx, AV_LOG_INFO, "Min level: %f\n", p->min);
214 av_log(ctx, AV_LOG_INFO, "Max level: %f\n", p->max);
215 av_log(ctx, AV_LOG_INFO, "Peak level dB: %f\n", LINEAR_TO_DB(FFMAX(-p->min, p->max)));
216 av_log(ctx, AV_LOG_INFO, "RMS level dB: %f\n", LINEAR_TO_DB(sqrt(p->sigma_x2 / p->nb_samples)));
217 av_log(ctx, AV_LOG_INFO, "RMS peak dB: %f\n", LINEAR_TO_DB(sqrt(p->max_sigma_x2)));
218 if (p->min_sigma_x2 != 1)
219 av_log(ctx, AV_LOG_INFO, "RMS trough dB: %f\n",LINEAR_TO_DB(sqrt(p->min_sigma_x2)));
220 av_log(ctx, AV_LOG_INFO, "Crest factor: %f\n", p->sigma_x2 ? FFMAX(-p->min, p->max) / sqrt(p->sigma_x2 / p->nb_samples) : 1);
221 av_log(ctx, AV_LOG_INFO, "Flat factor: %f\n", LINEAR_TO_DB((p->min_runs + p->max_runs) / (p->min_count + p->max_count)));
222 av_log(ctx, AV_LOG_INFO, "Peak count: %"PRId64"\n", p->min_count + p->max_count);
225 av_log(ctx, AV_LOG_INFO, "Overall\n");
226 av_log(ctx, AV_LOG_INFO, "DC offset: %f\n", max_sigma_x / (nb_samples / s->nb_channels));
227 av_log(ctx, AV_LOG_INFO, "Min level: %f\n", min);
228 av_log(ctx, AV_LOG_INFO, "Max level: %f\n", max);
229 av_log(ctx, AV_LOG_INFO, "Peak level dB: %f\n", LINEAR_TO_DB(FFMAX(-min, max)));
230 av_log(ctx, AV_LOG_INFO, "RMS level dB: %f\n", LINEAR_TO_DB(sqrt(sigma_x2 / nb_samples)));
231 av_log(ctx, AV_LOG_INFO, "RMS peak dB: %f\n", LINEAR_TO_DB(sqrt(max_sigma_x2)));
232 if (min_sigma_x2 != 1)
233 av_log(ctx, AV_LOG_INFO, "RMS trough dB: %f\n", LINEAR_TO_DB(sqrt(min_sigma_x2)));
234 av_log(ctx, AV_LOG_INFO, "Flat factor: %f\n", LINEAR_TO_DB((min_runs + max_runs) / (min_count + max_count)));
235 av_log(ctx, AV_LOG_INFO, "Peak count: %f\n", (min_count + max_count) / (double)s->nb_channels);
236 av_log(ctx, AV_LOG_INFO, "Number of samples: %"PRId64"\n", nb_samples / s->nb_channels);
239 static av_cold void uninit(AVFilterContext *ctx)
241 AudioStatsContext *s = ctx->priv;
244 av_freep(&s->chstats);
247 static const AVFilterPad astats_inputs[] = {
250 .type = AVMEDIA_TYPE_AUDIO,
251 .filter_frame = filter_frame,
256 static const AVFilterPad astats_outputs[] = {
259 .type = AVMEDIA_TYPE_AUDIO,
260 .config_props = config_output,
265 AVFilter avfilter_af_astats = {
267 .description = NULL_IF_CONFIG_SMALL("Show time domain statistics about audio frames."),
268 .query_formats = query_formats,
269 .priv_size = sizeof(AudioStatsContext),
270 .priv_class = &astats_class,
272 .inputs = astats_inputs,
273 .outputs = astats_outputs,