1 |
|
|
/* |
2 |
|
|
* Copyright (c) 2009 Rob Sykes <robs@users.sourceforge.net> |
3 |
|
|
* Copyright (c) 2013 Paul B Mahol |
4 |
|
|
* |
5 |
|
|
* This file is part of FFmpeg. |
6 |
|
|
* |
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. |
11 |
|
|
* |
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. |
16 |
|
|
* |
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 |
20 |
|
|
*/ |
21 |
|
|
|
22 |
|
|
#include <float.h> |
23 |
|
|
#include <math.h> |
24 |
|
|
|
25 |
|
|
#include "libavutil/opt.h" |
26 |
|
|
#include "audio.h" |
27 |
|
|
#include "avfilter.h" |
28 |
|
|
#include "internal.h" |
29 |
|
|
|
30 |
|
|
#define HISTOGRAM_SIZE 8192 |
31 |
|
|
#define HISTOGRAM_MAX (HISTOGRAM_SIZE-1) |
32 |
|
|
|
33 |
|
|
#define MEASURE_ALL UINT_MAX |
34 |
|
|
#define MEASURE_NONE 0 |
35 |
|
|
|
36 |
|
|
#define MEASURE_DC_OFFSET (1 << 0) |
37 |
|
|
#define MEASURE_MIN_LEVEL (1 << 1) |
38 |
|
|
#define MEASURE_MAX_LEVEL (1 << 2) |
39 |
|
|
#define MEASURE_MIN_DIFFERENCE (1 << 3) |
40 |
|
|
#define MEASURE_MAX_DIFFERENCE (1 << 4) |
41 |
|
|
#define MEASURE_MEAN_DIFFERENCE (1 << 5) |
42 |
|
|
#define MEASURE_RMS_DIFFERENCE (1 << 6) |
43 |
|
|
#define MEASURE_PEAK_LEVEL (1 << 7) |
44 |
|
|
#define MEASURE_RMS_LEVEL (1 << 8) |
45 |
|
|
#define MEASURE_RMS_PEAK (1 << 9) |
46 |
|
|
#define MEASURE_RMS_TROUGH (1 << 10) |
47 |
|
|
#define MEASURE_CREST_FACTOR (1 << 11) |
48 |
|
|
#define MEASURE_FLAT_FACTOR (1 << 12) |
49 |
|
|
#define MEASURE_PEAK_COUNT (1 << 13) |
50 |
|
|
#define MEASURE_BIT_DEPTH (1 << 14) |
51 |
|
|
#define MEASURE_DYNAMIC_RANGE (1 << 15) |
52 |
|
|
#define MEASURE_ZERO_CROSSINGS (1 << 16) |
53 |
|
|
#define MEASURE_ZERO_CROSSINGS_RATE (1 << 17) |
54 |
|
|
#define MEASURE_NUMBER_OF_SAMPLES (1 << 18) |
55 |
|
|
#define MEASURE_NUMBER_OF_NANS (1 << 19) |
56 |
|
|
#define MEASURE_NUMBER_OF_INFS (1 << 20) |
57 |
|
|
#define MEASURE_NUMBER_OF_DENORMALS (1 << 21) |
58 |
|
|
#define MEASURE_NOISE_FLOOR (1 << 22) |
59 |
|
|
#define MEASURE_NOISE_FLOOR_COUNT (1 << 23) |
60 |
|
|
|
61 |
|
|
#define MEASURE_MINMAXPEAK (MEASURE_MIN_LEVEL | MEASURE_MAX_LEVEL | MEASURE_PEAK_LEVEL) |
62 |
|
|
|
63 |
|
|
typedef struct ChannelStats { |
64 |
|
|
double last; |
65 |
|
|
double last_non_zero; |
66 |
|
|
double min_non_zero; |
67 |
|
|
double sigma_x, sigma_x2; |
68 |
|
|
double avg_sigma_x2, min_sigma_x2, max_sigma_x2; |
69 |
|
|
double min, max; |
70 |
|
|
double nmin, nmax; |
71 |
|
|
double min_run, max_run; |
72 |
|
|
double min_runs, max_runs; |
73 |
|
|
double min_diff, max_diff; |
74 |
|
|
double diff1_sum; |
75 |
|
|
double diff1_sum_x2; |
76 |
|
|
uint64_t mask, imask; |
77 |
|
|
uint64_t min_count, max_count; |
78 |
|
|
uint64_t noise_floor_count; |
79 |
|
|
uint64_t zero_runs; |
80 |
|
|
uint64_t nb_samples; |
81 |
|
|
uint64_t nb_nans; |
82 |
|
|
uint64_t nb_infs; |
83 |
|
|
uint64_t nb_denormals; |
84 |
|
|
double *win_samples; |
85 |
|
|
unsigned histogram[HISTOGRAM_SIZE]; |
86 |
|
|
int win_pos; |
87 |
|
|
int max_index; |
88 |
|
|
double noise_floor; |
89 |
|
|
} ChannelStats; |
90 |
|
|
|
91 |
|
|
typedef struct AudioStatsContext { |
92 |
|
|
const AVClass *class; |
93 |
|
|
ChannelStats *chstats; |
94 |
|
|
int nb_channels; |
95 |
|
|
uint64_t tc_samples; |
96 |
|
|
double time_constant; |
97 |
|
|
double mult; |
98 |
|
|
int metadata; |
99 |
|
|
int reset_count; |
100 |
|
|
int nb_frames; |
101 |
|
|
int maxbitdepth; |
102 |
|
|
int measure_perchannel; |
103 |
|
|
int measure_overall; |
104 |
|
|
int is_float; |
105 |
|
|
int is_double; |
106 |
|
|
} AudioStatsContext; |
107 |
|
|
|
108 |
|
|
#define OFFSET(x) offsetof(AudioStatsContext, x) |
109 |
|
|
#define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM |
110 |
|
|
|
111 |
|
|
static const AVOption astats_options[] = { |
112 |
|
|
{ "length", "set the window length", OFFSET(time_constant), AV_OPT_TYPE_DOUBLE, {.dbl=.05}, .01, 10, FLAGS }, |
113 |
|
|
{ "metadata", "inject metadata in the filtergraph", OFFSET(metadata), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS }, |
114 |
|
|
{ "reset", "recalculate stats after this many frames", OFFSET(reset_count), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS }, |
115 |
|
|
{ "measure_perchannel", "only measure_perchannel these per-channel statistics", OFFSET(measure_perchannel), AV_OPT_TYPE_FLAGS, {.i64=MEASURE_ALL}, 0, UINT_MAX, FLAGS, "measure" }, |
116 |
|
|
{ "none" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NONE }, 0, 0, FLAGS, "measure" }, |
117 |
|
|
{ "all" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_ALL }, 0, 0, FLAGS, "measure" }, |
118 |
|
|
{ "DC_offset" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_DC_OFFSET }, 0, 0, FLAGS, "measure" }, |
119 |
|
|
{ "Min_level" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_MIN_LEVEL }, 0, 0, FLAGS, "measure" }, |
120 |
|
|
{ "Max_level" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_MAX_LEVEL }, 0, 0, FLAGS, "measure" }, |
121 |
|
|
{ "Min_difference" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_MIN_DIFFERENCE }, 0, 0, FLAGS, "measure" }, |
122 |
|
|
{ "Max_difference" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_MAX_DIFFERENCE }, 0, 0, FLAGS, "measure" }, |
123 |
|
|
{ "Mean_difference" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_MEAN_DIFFERENCE }, 0, 0, FLAGS, "measure" }, |
124 |
|
|
{ "RMS_difference" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_RMS_DIFFERENCE }, 0, 0, FLAGS, "measure" }, |
125 |
|
|
{ "Peak_level" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_PEAK_LEVEL }, 0, 0, FLAGS, "measure" }, |
126 |
|
|
{ "RMS_level" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_RMS_LEVEL }, 0, 0, FLAGS, "measure" }, |
127 |
|
|
{ "RMS_peak" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_RMS_PEAK }, 0, 0, FLAGS, "measure" }, |
128 |
|
|
{ "RMS_trough" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_RMS_TROUGH }, 0, 0, FLAGS, "measure" }, |
129 |
|
|
{ "Crest_factor" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_CREST_FACTOR }, 0, 0, FLAGS, "measure" }, |
130 |
|
|
{ "Flat_factor" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_FLAT_FACTOR }, 0, 0, FLAGS, "measure" }, |
131 |
|
|
{ "Peak_count" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_PEAK_COUNT }, 0, 0, FLAGS, "measure" }, |
132 |
|
|
{ "Bit_depth" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_BIT_DEPTH }, 0, 0, FLAGS, "measure" }, |
133 |
|
|
{ "Dynamic_range" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_DYNAMIC_RANGE }, 0, 0, FLAGS, "measure" }, |
134 |
|
|
{ "Zero_crossings" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_ZERO_CROSSINGS }, 0, 0, FLAGS, "measure" }, |
135 |
|
|
{ "Zero_crossings_rate" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_ZERO_CROSSINGS_RATE }, 0, 0, FLAGS, "measure" }, |
136 |
|
|
{ "Noise_floor" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NOISE_FLOOR }, 0, 0, FLAGS, "measure" }, |
137 |
|
|
{ "Noise_floor_count" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NOISE_FLOOR_COUNT }, 0, 0, FLAGS, "measure" }, |
138 |
|
|
{ "Number_of_samples" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NUMBER_OF_SAMPLES }, 0, 0, FLAGS, "measure" }, |
139 |
|
|
{ "Number_of_NaNs" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NUMBER_OF_NANS }, 0, 0, FLAGS, "measure" }, |
140 |
|
|
{ "Number_of_Infs" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NUMBER_OF_INFS }, 0, 0, FLAGS, "measure" }, |
141 |
|
|
{ "Number_of_denormals" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NUMBER_OF_DENORMALS }, 0, 0, FLAGS, "measure" }, |
142 |
|
|
{ "measure_overall", "only measure_perchannel these overall statistics", OFFSET(measure_overall), AV_OPT_TYPE_FLAGS, {.i64=MEASURE_ALL}, 0, UINT_MAX, FLAGS, "measure" }, |
143 |
|
|
{ NULL } |
144 |
|
|
}; |
145 |
|
|
|
146 |
|
|
AVFILTER_DEFINE_CLASS(astats); |
147 |
|
|
|
148 |
|
|
static int query_formats(AVFilterContext *ctx) |
149 |
|
|
{ |
150 |
|
|
AVFilterFormats *formats; |
151 |
|
|
AVFilterChannelLayouts *layouts; |
152 |
|
|
static const enum AVSampleFormat sample_fmts[] = { |
153 |
|
|
AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S16P, |
154 |
|
|
AV_SAMPLE_FMT_S32, AV_SAMPLE_FMT_S32P, |
155 |
|
|
AV_SAMPLE_FMT_S64, AV_SAMPLE_FMT_S64P, |
156 |
|
|
AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_FLTP, |
157 |
|
|
AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_DBLP, |
158 |
|
|
AV_SAMPLE_FMT_NONE |
159 |
|
|
}; |
160 |
|
|
int ret; |
161 |
|
|
|
162 |
|
|
layouts = ff_all_channel_counts(); |
163 |
|
|
if (!layouts) |
164 |
|
|
return AVERROR(ENOMEM); |
165 |
|
|
ret = ff_set_common_channel_layouts(ctx, layouts); |
166 |
|
|
if (ret < 0) |
167 |
|
|
return ret; |
168 |
|
|
|
169 |
|
|
formats = ff_make_format_list(sample_fmts); |
170 |
|
|
if (!formats) |
171 |
|
|
return AVERROR(ENOMEM); |
172 |
|
|
ret = ff_set_common_formats(ctx, formats); |
173 |
|
|
if (ret < 0) |
174 |
|
|
return ret; |
175 |
|
|
|
176 |
|
|
formats = ff_all_samplerates(); |
177 |
|
|
if (!formats) |
178 |
|
|
return AVERROR(ENOMEM); |
179 |
|
|
return ff_set_common_samplerates(ctx, formats); |
180 |
|
|
} |
181 |
|
|
|
182 |
|
|
static void reset_stats(AudioStatsContext *s) |
183 |
|
|
{ |
184 |
|
|
int c; |
185 |
|
|
|
186 |
|
|
for (c = 0; c < s->nb_channels; c++) { |
187 |
|
|
ChannelStats *p = &s->chstats[c]; |
188 |
|
|
|
189 |
|
|
p->min = p->nmin = p->min_sigma_x2 = DBL_MAX; |
190 |
|
|
p->max = p->nmax = p->max_sigma_x2 =-DBL_MAX; |
191 |
|
|
p->min_non_zero = DBL_MAX; |
192 |
|
|
p->min_diff = DBL_MAX; |
193 |
|
|
p->max_diff = 0; |
194 |
|
|
p->sigma_x = 0; |
195 |
|
|
p->sigma_x2 = 0; |
196 |
|
|
p->avg_sigma_x2 = 0; |
197 |
|
|
p->min_run = 0; |
198 |
|
|
p->max_run = 0; |
199 |
|
|
p->min_runs = 0; |
200 |
|
|
p->max_runs = 0; |
201 |
|
|
p->diff1_sum = 0; |
202 |
|
|
p->diff1_sum_x2 = 0; |
203 |
|
|
p->mask = 0; |
204 |
|
|
p->imask = 0xFFFFFFFFFFFFFFFF; |
205 |
|
|
p->min_count = 0; |
206 |
|
|
p->max_count = 0; |
207 |
|
|
p->zero_runs = 0; |
208 |
|
|
p->nb_samples = 0; |
209 |
|
|
p->nb_nans = 0; |
210 |
|
|
p->nb_infs = 0; |
211 |
|
|
p->nb_denormals = 0; |
212 |
|
|
p->last = NAN; |
213 |
|
|
p->noise_floor = NAN; |
214 |
|
|
p->noise_floor_count = 0; |
215 |
|
|
p->win_pos = 0; |
216 |
|
|
memset(p->win_samples, 0, s->tc_samples * sizeof(*p->win_samples)); |
217 |
|
|
memset(p->histogram, 0, sizeof(p->histogram)); |
218 |
|
|
} |
219 |
|
|
} |
220 |
|
|
|
221 |
|
|
static int config_output(AVFilterLink *outlink) |
222 |
|
|
{ |
223 |
|
|
AudioStatsContext *s = outlink->src->priv; |
224 |
|
|
|
225 |
|
|
s->chstats = av_calloc(sizeof(*s->chstats), outlink->channels); |
226 |
|
|
if (!s->chstats) |
227 |
|
|
return AVERROR(ENOMEM); |
228 |
|
|
|
229 |
|
|
s->tc_samples = 5 * s->time_constant * outlink->sample_rate + .5; |
230 |
|
|
s->nb_channels = outlink->channels; |
231 |
|
|
|
232 |
|
|
for (int i = 0; i < s->nb_channels; i++) { |
233 |
|
|
ChannelStats *p = &s->chstats[i]; |
234 |
|
|
|
235 |
|
|
p->win_samples = av_calloc(s->tc_samples, sizeof(*p->win_samples)); |
236 |
|
|
if (!p->win_samples) |
237 |
|
|
return AVERROR(ENOMEM); |
238 |
|
|
} |
239 |
|
|
|
240 |
|
|
s->mult = exp((-1 / s->time_constant / outlink->sample_rate)); |
241 |
|
|
s->nb_frames = 0; |
242 |
|
|
s->maxbitdepth = av_get_bytes_per_sample(outlink->format) * 8; |
243 |
|
|
s->is_double = outlink->format == AV_SAMPLE_FMT_DBL || |
244 |
|
|
outlink->format == AV_SAMPLE_FMT_DBLP; |
245 |
|
|
|
246 |
|
|
s->is_float = outlink->format == AV_SAMPLE_FMT_FLT || |
247 |
|
|
outlink->format == AV_SAMPLE_FMT_FLTP; |
248 |
|
|
|
249 |
|
|
reset_stats(s); |
250 |
|
|
|
251 |
|
|
return 0; |
252 |
|
|
} |
253 |
|
|
|
254 |
|
|
static void bit_depth(AudioStatsContext *s, uint64_t mask, uint64_t imask, AVRational *depth) |
255 |
|
|
{ |
256 |
|
|
unsigned result = s->maxbitdepth; |
257 |
|
|
|
258 |
|
|
mask = mask & (~imask); |
259 |
|
|
|
260 |
|
|
for (; result && !(mask & 1); --result, mask >>= 1); |
261 |
|
|
|
262 |
|
|
depth->den = result; |
263 |
|
|
depth->num = 0; |
264 |
|
|
|
265 |
|
|
for (; result; --result, mask >>= 1) |
266 |
|
|
if (mask & 1) |
267 |
|
|
depth->num++; |
268 |
|
|
} |
269 |
|
|
|
270 |
|
|
static inline void update_minmax(AudioStatsContext *s, ChannelStats *p, double d) |
271 |
|
|
{ |
272 |
|
|
if (d < p->min) |
273 |
|
|
p->min = d; |
274 |
|
|
if (d > p->max) |
275 |
|
|
p->max = d; |
276 |
|
|
} |
277 |
|
|
|
278 |
|
|
static inline void update_stat(AudioStatsContext *s, ChannelStats *p, double d, double nd, int64_t i) |
279 |
|
|
{ |
280 |
|
|
double drop; |
281 |
|
|
int index; |
282 |
|
|
|
283 |
|
|
if (d < p->min) { |
284 |
|
|
p->min = d; |
285 |
|
|
p->nmin = nd; |
286 |
|
|
p->min_run = 1; |
287 |
|
|
p->min_runs = 0; |
288 |
|
|
p->min_count = 1; |
289 |
|
|
} else if (d == p->min) { |
290 |
|
|
p->min_count++; |
291 |
|
|
p->min_run = d == p->last ? p->min_run + 1 : 1; |
292 |
|
|
} else if (p->last == p->min) { |
293 |
|
|
p->min_runs += p->min_run * p->min_run; |
294 |
|
|
} |
295 |
|
|
|
296 |
|
|
if (d != 0 && FFABS(d) < p->min_non_zero) |
297 |
|
|
p->min_non_zero = FFABS(d); |
298 |
|
|
|
299 |
|
|
if (d > p->max) { |
300 |
|
|
p->max = d; |
301 |
|
|
p->nmax = nd; |
302 |
|
|
p->max_run = 1; |
303 |
|
|
p->max_runs = 0; |
304 |
|
|
p->max_count = 1; |
305 |
|
|
} else if (d == p->max) { |
306 |
|
|
p->max_count++; |
307 |
|
|
p->max_run = d == p->last ? p->max_run + 1 : 1; |
308 |
|
|
} else if (p->last == p->max) { |
309 |
|
|
p->max_runs += p->max_run * p->max_run; |
310 |
|
|
} |
311 |
|
|
|
312 |
|
|
if (d != 0) { |
313 |
|
|
p->zero_runs += FFSIGN(d) != FFSIGN(p->last_non_zero); |
314 |
|
|
p->last_non_zero = d; |
315 |
|
|
} |
316 |
|
|
|
317 |
|
|
p->sigma_x += nd; |
318 |
|
|
p->sigma_x2 += nd * nd; |
319 |
|
|
p->avg_sigma_x2 = p->avg_sigma_x2 * s->mult + (1.0 - s->mult) * nd * nd; |
320 |
|
|
if (!isnan(p->last)) { |
321 |
|
|
p->min_diff = FFMIN(p->min_diff, fabs(d - p->last)); |
322 |
|
|
p->max_diff = FFMAX(p->max_diff, fabs(d - p->last)); |
323 |
|
|
p->diff1_sum += fabs(d - p->last); |
324 |
|
|
p->diff1_sum_x2 += (d - p->last) * (d - p->last); |
325 |
|
|
} |
326 |
|
|
p->last = d; |
327 |
|
|
p->mask |= i; |
328 |
|
|
p->imask &= i; |
329 |
|
|
|
330 |
|
|
drop = p->win_samples[p->win_pos]; |
331 |
|
|
p->win_samples[p->win_pos] = nd; |
332 |
|
|
index = av_clip(lrint(av_clipd(FFABS(nd), 0.0, 1.0) * HISTOGRAM_MAX), 0, HISTOGRAM_MAX); |
333 |
|
|
p->max_index = FFMAX(p->max_index, index); |
334 |
|
|
p->histogram[index]++; |
335 |
|
|
if (!isnan(p->noise_floor)) |
336 |
|
|
p->histogram[av_clip(lrint(av_clipd(FFABS(drop), 0.0, 1.0) * HISTOGRAM_MAX), 0, HISTOGRAM_MAX)]--; |
337 |
|
|
p->win_pos++; |
338 |
|
|
|
339 |
|
|
while (p->histogram[p->max_index] == 0) |
340 |
|
|
p->max_index--; |
341 |
|
|
if (p->win_pos >= s->tc_samples || !isnan(p->noise_floor)) { |
342 |
|
|
double noise_floor = 1.; |
343 |
|
|
|
344 |
|
|
for (int i = p->max_index; i >= 0; i--) { |
345 |
|
|
if (p->histogram[i]) { |
346 |
|
|
noise_floor = i / (double)HISTOGRAM_MAX; |
347 |
|
|
break; |
348 |
|
|
} |
349 |
|
|
} |
350 |
|
|
|
351 |
|
|
if (isnan(p->noise_floor)) { |
352 |
|
|
p->noise_floor = noise_floor; |
353 |
|
|
p->noise_floor_count = 1; |
354 |
|
|
} else { |
355 |
|
|
if (noise_floor < p->noise_floor) { |
356 |
|
|
p->noise_floor = noise_floor; |
357 |
|
|
p->noise_floor_count = 1; |
358 |
|
|
} else if (noise_floor == p->noise_floor) { |
359 |
|
|
p->noise_floor_count++; |
360 |
|
|
} |
361 |
|
|
} |
362 |
|
|
} |
363 |
|
|
|
364 |
|
|
if (p->win_pos >= s->tc_samples) { |
365 |
|
|
p->win_pos = 0; |
366 |
|
|
} |
367 |
|
|
|
368 |
|
|
if (p->nb_samples >= s->tc_samples) { |
369 |
|
|
p->max_sigma_x2 = FFMAX(p->max_sigma_x2, p->avg_sigma_x2); |
370 |
|
|
p->min_sigma_x2 = FFMIN(p->min_sigma_x2, p->avg_sigma_x2); |
371 |
|
|
} |
372 |
|
|
p->nb_samples++; |
373 |
|
|
} |
374 |
|
|
|
375 |
|
|
static inline void update_float_stat(AudioStatsContext *s, ChannelStats *p, float d) |
376 |
|
|
{ |
377 |
|
|
int type = fpclassify(d); |
378 |
|
|
|
379 |
|
|
p->nb_nans += type == FP_NAN; |
380 |
|
|
p->nb_infs += type == FP_INFINITE; |
381 |
|
|
p->nb_denormals += type == FP_SUBNORMAL; |
382 |
|
|
} |
383 |
|
|
|
384 |
|
|
static inline void update_double_stat(AudioStatsContext *s, ChannelStats *p, double d) |
385 |
|
|
{ |
386 |
|
|
int type = fpclassify(d); |
387 |
|
|
|
388 |
|
|
p->nb_nans += type == FP_NAN; |
389 |
|
|
p->nb_infs += type == FP_INFINITE; |
390 |
|
|
p->nb_denormals += type == FP_SUBNORMAL; |
391 |
|
|
} |
392 |
|
|
|
393 |
|
|
static void set_meta(AVDictionary **metadata, int chan, const char *key, |
394 |
|
|
const char *fmt, double val) |
395 |
|
|
{ |
396 |
|
|
uint8_t value[128]; |
397 |
|
|
uint8_t key2[128]; |
398 |
|
|
|
399 |
|
|
snprintf(value, sizeof(value), fmt, val); |
400 |
|
|
if (chan) |
401 |
|
|
snprintf(key2, sizeof(key2), "lavfi.astats.%d.%s", chan, key); |
402 |
|
|
else |
403 |
|
|
snprintf(key2, sizeof(key2), "lavfi.astats.%s", key); |
404 |
|
|
av_dict_set(metadata, key2, value, 0); |
405 |
|
|
} |
406 |
|
|
|
407 |
|
|
#define LINEAR_TO_DB(x) (log10(x) * 20) |
408 |
|
|
|
409 |
|
|
static void set_metadata(AudioStatsContext *s, AVDictionary **metadata) |
410 |
|
|
{ |
411 |
|
|
uint64_t mask = 0, imask = 0xFFFFFFFFFFFFFFFF, min_count = 0, max_count = 0, nb_samples = 0, noise_floor_count = 0; |
412 |
|
|
uint64_t nb_nans = 0, nb_infs = 0, nb_denormals = 0; |
413 |
|
|
double min_runs = 0, max_runs = 0, |
414 |
|
|
min = DBL_MAX, max =-DBL_MAX, min_diff = DBL_MAX, max_diff = 0, |
415 |
|
|
nmin = DBL_MAX, nmax =-DBL_MAX, |
416 |
|
|
max_sigma_x = 0, |
417 |
|
|
diff1_sum = 0, |
418 |
|
|
diff1_sum_x2 = 0, |
419 |
|
|
sigma_x = 0, |
420 |
|
|
sigma_x2 = 0, |
421 |
|
|
noise_floor = 0, |
422 |
|
|
min_sigma_x2 = DBL_MAX, |
423 |
|
|
max_sigma_x2 =-DBL_MAX; |
424 |
|
|
AVRational depth; |
425 |
|
|
int c; |
426 |
|
|
|
427 |
|
|
for (c = 0; c < s->nb_channels; c++) { |
428 |
|
|
ChannelStats *p = &s->chstats[c]; |
429 |
|
|
|
430 |
|
|
if (p->nb_samples < s->tc_samples) |
431 |
|
|
p->min_sigma_x2 = p->max_sigma_x2 = p->sigma_x2 / p->nb_samples; |
432 |
|
|
|
433 |
|
|
min = FFMIN(min, p->min); |
434 |
|
|
max = FFMAX(max, p->max); |
435 |
|
|
nmin = FFMIN(nmin, p->nmin); |
436 |
|
|
nmax = FFMAX(nmax, p->nmax); |
437 |
|
|
min_diff = FFMIN(min_diff, p->min_diff); |
438 |
|
|
max_diff = FFMAX(max_diff, p->max_diff); |
439 |
|
|
diff1_sum += p->diff1_sum; |
440 |
|
|
diff1_sum_x2 += p->diff1_sum_x2; |
441 |
|
|
min_sigma_x2 = FFMIN(min_sigma_x2, p->min_sigma_x2); |
442 |
|
|
max_sigma_x2 = FFMAX(max_sigma_x2, p->max_sigma_x2); |
443 |
|
|
sigma_x += p->sigma_x; |
444 |
|
|
sigma_x2 += p->sigma_x2; |
445 |
|
|
noise_floor = FFMAX(noise_floor, p->noise_floor); |
446 |
|
|
noise_floor_count += p->noise_floor_count; |
447 |
|
|
min_count += p->min_count; |
448 |
|
|
max_count += p->max_count; |
449 |
|
|
min_runs += p->min_runs; |
450 |
|
|
max_runs += p->max_runs; |
451 |
|
|
mask |= p->mask; |
452 |
|
|
imask &= p->imask; |
453 |
|
|
nb_samples += p->nb_samples; |
454 |
|
|
nb_nans += p->nb_nans; |
455 |
|
|
nb_infs += p->nb_infs; |
456 |
|
|
nb_denormals += p->nb_denormals; |
457 |
|
|
if (fabs(p->sigma_x) > fabs(max_sigma_x)) |
458 |
|
|
max_sigma_x = p->sigma_x; |
459 |
|
|
|
460 |
|
|
if (s->measure_perchannel & MEASURE_DC_OFFSET) |
461 |
|
|
set_meta(metadata, c + 1, "DC_offset", "%f", p->sigma_x / p->nb_samples); |
462 |
|
|
if (s->measure_perchannel & MEASURE_MIN_LEVEL) |
463 |
|
|
set_meta(metadata, c + 1, "Min_level", "%f", p->min); |
464 |
|
|
if (s->measure_perchannel & MEASURE_MAX_LEVEL) |
465 |
|
|
set_meta(metadata, c + 1, "Max_level", "%f", p->max); |
466 |
|
|
if (s->measure_perchannel & MEASURE_MIN_DIFFERENCE) |
467 |
|
|
set_meta(metadata, c + 1, "Min_difference", "%f", p->min_diff); |
468 |
|
|
if (s->measure_perchannel & MEASURE_MAX_DIFFERENCE) |
469 |
|
|
set_meta(metadata, c + 1, "Max_difference", "%f", p->max_diff); |
470 |
|
|
if (s->measure_perchannel & MEASURE_MEAN_DIFFERENCE) |
471 |
|
|
set_meta(metadata, c + 1, "Mean_difference", "%f", p->diff1_sum / (p->nb_samples - 1)); |
472 |
|
|
if (s->measure_perchannel & MEASURE_RMS_DIFFERENCE) |
473 |
|
|
set_meta(metadata, c + 1, "RMS_difference", "%f", sqrt(p->diff1_sum_x2 / (p->nb_samples - 1))); |
474 |
|
|
if (s->measure_perchannel & MEASURE_PEAK_LEVEL) |
475 |
|
|
set_meta(metadata, c + 1, "Peak_level", "%f", LINEAR_TO_DB(FFMAX(-p->nmin, p->nmax))); |
476 |
|
|
if (s->measure_perchannel & MEASURE_RMS_LEVEL) |
477 |
|
|
set_meta(metadata, c + 1, "RMS_level", "%f", LINEAR_TO_DB(sqrt(p->sigma_x2 / p->nb_samples))); |
478 |
|
|
if (s->measure_perchannel & MEASURE_RMS_PEAK) |
479 |
|
|
set_meta(metadata, c + 1, "RMS_peak", "%f", LINEAR_TO_DB(sqrt(p->max_sigma_x2))); |
480 |
|
|
if (s->measure_perchannel & MEASURE_RMS_TROUGH) |
481 |
|
|
set_meta(metadata, c + 1, "RMS_trough", "%f", LINEAR_TO_DB(sqrt(p->min_sigma_x2))); |
482 |
|
|
if (s->measure_perchannel & MEASURE_CREST_FACTOR) |
483 |
|
|
set_meta(metadata, c + 1, "Crest_factor", "%f", p->sigma_x2 ? FFMAX(-p->min, p->max) / sqrt(p->sigma_x2 / p->nb_samples) : 1); |
484 |
|
|
if (s->measure_perchannel & MEASURE_FLAT_FACTOR) |
485 |
|
|
set_meta(metadata, c + 1, "Flat_factor", "%f", LINEAR_TO_DB((p->min_runs + p->max_runs) / (p->min_count + p->max_count))); |
486 |
|
|
if (s->measure_perchannel & MEASURE_PEAK_COUNT) |
487 |
|
|
set_meta(metadata, c + 1, "Peak_count", "%f", (float)(p->min_count + p->max_count)); |
488 |
|
|
if (s->measure_perchannel & MEASURE_NOISE_FLOOR) |
489 |
|
|
set_meta(metadata, c + 1, "Noise_floor", "%f", LINEAR_TO_DB(p->noise_floor)); |
490 |
|
|
if (s->measure_perchannel & MEASURE_NOISE_FLOOR_COUNT) |
491 |
|
|
set_meta(metadata, c + 1, "Noise_floor_count", "%f", p->noise_floor_count); |
492 |
|
|
if (s->measure_perchannel & MEASURE_BIT_DEPTH) { |
493 |
|
|
bit_depth(s, p->mask, p->imask, &depth); |
494 |
|
|
set_meta(metadata, c + 1, "Bit_depth", "%f", depth.num); |
495 |
|
|
set_meta(metadata, c + 1, "Bit_depth2", "%f", depth.den); |
496 |
|
|
} |
497 |
|
|
if (s->measure_perchannel & MEASURE_DYNAMIC_RANGE) |
498 |
|
|
set_meta(metadata, c + 1, "Dynamic_range", "%f", LINEAR_TO_DB(2 * FFMAX(FFABS(p->min), FFABS(p->max))/ p->min_non_zero)); |
499 |
|
|
if (s->measure_perchannel & MEASURE_ZERO_CROSSINGS) |
500 |
|
|
set_meta(metadata, c + 1, "Zero_crossings", "%f", p->zero_runs); |
501 |
|
|
if (s->measure_perchannel & MEASURE_ZERO_CROSSINGS_RATE) |
502 |
|
|
set_meta(metadata, c + 1, "Zero_crossings_rate", "%f", p->zero_runs/(double)p->nb_samples); |
503 |
|
|
if ((s->is_float || s->is_double) && s->measure_perchannel & MEASURE_NUMBER_OF_NANS) |
504 |
|
|
set_meta(metadata, c + 1, "Number of NaNs", "%f", p->nb_nans); |
505 |
|
|
if ((s->is_float || s->is_double) && s->measure_perchannel & MEASURE_NUMBER_OF_INFS) |
506 |
|
|
set_meta(metadata, c + 1, "Number of Infs", "%f", p->nb_infs); |
507 |
|
|
if ((s->is_float || s->is_double) && s->measure_perchannel & MEASURE_NUMBER_OF_DENORMALS) |
508 |
|
|
set_meta(metadata, c + 1, "Number of denormals", "%f", p->nb_denormals); |
509 |
|
|
} |
510 |
|
|
|
511 |
|
|
if (s->measure_overall & MEASURE_DC_OFFSET) |
512 |
|
|
set_meta(metadata, 0, "Overall.DC_offset", "%f", max_sigma_x / (nb_samples / s->nb_channels)); |
513 |
|
|
if (s->measure_overall & MEASURE_MIN_LEVEL) |
514 |
|
|
set_meta(metadata, 0, "Overall.Min_level", "%f", min); |
515 |
|
|
if (s->measure_overall & MEASURE_MAX_LEVEL) |
516 |
|
|
set_meta(metadata, 0, "Overall.Max_level", "%f", max); |
517 |
|
|
if (s->measure_overall & MEASURE_MIN_DIFFERENCE) |
518 |
|
|
set_meta(metadata, 0, "Overall.Min_difference", "%f", min_diff); |
519 |
|
|
if (s->measure_overall & MEASURE_MAX_DIFFERENCE) |
520 |
|
|
set_meta(metadata, 0, "Overall.Max_difference", "%f", max_diff); |
521 |
|
|
if (s->measure_overall & MEASURE_MEAN_DIFFERENCE) |
522 |
|
|
set_meta(metadata, 0, "Overall.Mean_difference", "%f", diff1_sum / (nb_samples - s->nb_channels)); |
523 |
|
|
if (s->measure_overall & MEASURE_RMS_DIFFERENCE) |
524 |
|
|
set_meta(metadata, 0, "Overall.RMS_difference", "%f", sqrt(diff1_sum_x2 / (nb_samples - s->nb_channels))); |
525 |
|
|
if (s->measure_overall & MEASURE_PEAK_LEVEL) |
526 |
|
|
set_meta(metadata, 0, "Overall.Peak_level", "%f", LINEAR_TO_DB(FFMAX(-nmin, nmax))); |
527 |
|
|
if (s->measure_overall & MEASURE_RMS_LEVEL) |
528 |
|
|
set_meta(metadata, 0, "Overall.RMS_level", "%f", LINEAR_TO_DB(sqrt(sigma_x2 / nb_samples))); |
529 |
|
|
if (s->measure_overall & MEASURE_RMS_PEAK) |
530 |
|
|
set_meta(metadata, 0, "Overall.RMS_peak", "%f", LINEAR_TO_DB(sqrt(max_sigma_x2))); |
531 |
|
|
if (s->measure_overall & MEASURE_RMS_TROUGH) |
532 |
|
|
set_meta(metadata, 0, "Overall.RMS_trough", "%f", LINEAR_TO_DB(sqrt(min_sigma_x2))); |
533 |
|
|
if (s->measure_overall & MEASURE_FLAT_FACTOR) |
534 |
|
|
set_meta(metadata, 0, "Overall.Flat_factor", "%f", LINEAR_TO_DB((min_runs + max_runs) / (min_count + max_count))); |
535 |
|
|
if (s->measure_overall & MEASURE_PEAK_COUNT) |
536 |
|
|
set_meta(metadata, 0, "Overall.Peak_count", "%f", (float)(min_count + max_count) / (double)s->nb_channels); |
537 |
|
|
if (s->measure_overall & MEASURE_NOISE_FLOOR) |
538 |
|
|
set_meta(metadata, 0, "Overall.Noise_floor", "%f", LINEAR_TO_DB(noise_floor)); |
539 |
|
|
if (s->measure_overall & MEASURE_NOISE_FLOOR_COUNT) |
540 |
|
|
set_meta(metadata, 0, "Overall.Noise_floor_count", "%f", noise_floor_count / (double)s->nb_channels); |
541 |
|
|
if (s->measure_overall & MEASURE_BIT_DEPTH) { |
542 |
|
|
bit_depth(s, mask, imask, &depth); |
543 |
|
|
set_meta(metadata, 0, "Overall.Bit_depth", "%f", depth.num); |
544 |
|
|
set_meta(metadata, 0, "Overall.Bit_depth2", "%f", depth.den); |
545 |
|
|
} |
546 |
|
|
if (s->measure_overall & MEASURE_NUMBER_OF_SAMPLES) |
547 |
|
|
set_meta(metadata, 0, "Overall.Number_of_samples", "%f", nb_samples / s->nb_channels); |
548 |
|
|
if ((s->is_float || s->is_double) && s->measure_overall & MEASURE_NUMBER_OF_NANS) |
549 |
|
|
set_meta(metadata, 0, "Number of NaNs", "%f", nb_nans / (float)s->nb_channels); |
550 |
|
|
if ((s->is_float || s->is_double) && s->measure_overall & MEASURE_NUMBER_OF_INFS) |
551 |
|
|
set_meta(metadata, 0, "Number of Infs", "%f", nb_infs / (float)s->nb_channels); |
552 |
|
|
if ((s->is_float || s->is_double) && s->measure_overall & MEASURE_NUMBER_OF_DENORMALS) |
553 |
|
|
set_meta(metadata, 0, "Number of denormals", "%f", nb_denormals / (float)s->nb_channels); |
554 |
|
|
} |
555 |
|
|
|
556 |
|
|
#define UPDATE_STATS_P(type, update_func, update_float, channel_func) \ |
557 |
|
|
for (int c = start; c < end; c++) { \ |
558 |
|
|
ChannelStats *p = &s->chstats[c]; \ |
559 |
|
|
const type *src = (const type *)data[c]; \ |
560 |
|
|
const type * const srcend = src + samples; \ |
561 |
|
|
for (; src < srcend; src++) { \ |
562 |
|
|
update_func; \ |
563 |
|
|
update_float; \ |
564 |
|
|
} \ |
565 |
|
|
channel_func; \ |
566 |
|
|
} |
567 |
|
|
|
568 |
|
|
#define UPDATE_STATS_I(type, update_func, update_float, channel_func) \ |
569 |
|
|
for (int c = start; c < end; c++) { \ |
570 |
|
|
ChannelStats *p = &s->chstats[c]; \ |
571 |
|
|
const type *src = (const type *)data[0]; \ |
572 |
|
|
const type * const srcend = src + samples * channels; \ |
573 |
|
|
for (src += c; src < srcend; src += channels) { \ |
574 |
|
|
update_func; \ |
575 |
|
|
update_float; \ |
576 |
|
|
} \ |
577 |
|
|
channel_func; \ |
578 |
|
|
} |
579 |
|
|
|
580 |
|
|
#define UPDATE_STATS(planar, type, sample, normalizer_suffix, int_sample) \ |
581 |
|
|
if ((s->measure_overall | s->measure_perchannel) & ~MEASURE_MINMAXPEAK) { \ |
582 |
|
|
UPDATE_STATS_##planar(type, update_stat(s, p, sample, sample normalizer_suffix, int_sample), s->is_float ? update_float_stat(s, p, sample) : s->is_double ? update_double_stat(s, p, sample) : (void)NULL, ); \ |
583 |
|
|
} else { \ |
584 |
|
|
UPDATE_STATS_##planar(type, update_minmax(s, p, sample), , p->nmin = p->min normalizer_suffix; p->nmax = p->max normalizer_suffix;); \ |
585 |
|
|
} |
586 |
|
|
|
587 |
|
|
static int filter_channel(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) |
588 |
|
|
{ |
589 |
|
|
AudioStatsContext *s = ctx->priv; |
590 |
|
|
AVFilterLink *inlink = ctx->inputs[0]; |
591 |
|
|
AVFrame *buf = arg; |
592 |
|
|
const uint8_t * const * const data = (const uint8_t * const *)buf->extended_data; |
593 |
|
|
const int channels = s->nb_channels; |
594 |
|
|
const int samples = buf->nb_samples; |
595 |
|
|
const int start = (buf->channels * jobnr) / nb_jobs; |
596 |
|
|
const int end = (buf->channels * (jobnr+1)) / nb_jobs; |
597 |
|
|
|
598 |
|
|
switch (inlink->format) { |
599 |
|
|
case AV_SAMPLE_FMT_DBLP: |
600 |
|
|
UPDATE_STATS(P, double, *src, , llrint(*src * (UINT64_C(1) << 63))); |
601 |
|
|
break; |
602 |
|
|
case AV_SAMPLE_FMT_DBL: |
603 |
|
|
UPDATE_STATS(I, double, *src, , llrint(*src * (UINT64_C(1) << 63))); |
604 |
|
|
break; |
605 |
|
|
case AV_SAMPLE_FMT_FLTP: |
606 |
|
|
UPDATE_STATS(P, float, *src, , llrint(*src * (UINT64_C(1) << 31))); |
607 |
|
|
break; |
608 |
|
|
case AV_SAMPLE_FMT_FLT: |
609 |
|
|
UPDATE_STATS(I, float, *src, , llrint(*src * (UINT64_C(1) << 31))); |
610 |
|
|
break; |
611 |
|
|
case AV_SAMPLE_FMT_S64P: |
612 |
|
|
UPDATE_STATS(P, int64_t, *src, / (double)INT64_MAX, *src); |
613 |
|
|
break; |
614 |
|
|
case AV_SAMPLE_FMT_S64: |
615 |
|
|
UPDATE_STATS(I, int64_t, *src, / (double)INT64_MAX, *src); |
616 |
|
|
break; |
617 |
|
|
case AV_SAMPLE_FMT_S32P: |
618 |
|
|
UPDATE_STATS(P, int32_t, *src, / (double)INT32_MAX, *src); |
619 |
|
|
break; |
620 |
|
|
case AV_SAMPLE_FMT_S32: |
621 |
|
|
UPDATE_STATS(I, int32_t, *src, / (double)INT32_MAX, *src); |
622 |
|
|
break; |
623 |
|
|
case AV_SAMPLE_FMT_S16P: |
624 |
|
|
UPDATE_STATS(P, int16_t, *src, / (double)INT16_MAX, *src); |
625 |
|
|
break; |
626 |
|
|
case AV_SAMPLE_FMT_S16: |
627 |
|
|
UPDATE_STATS(I, int16_t, *src, / (double)INT16_MAX, *src); |
628 |
|
|
break; |
629 |
|
|
} |
630 |
|
|
|
631 |
|
|
return 0; |
632 |
|
|
} |
633 |
|
|
|
634 |
|
|
static int filter_frame(AVFilterLink *inlink, AVFrame *buf) |
635 |
|
|
{ |
636 |
|
|
AVFilterContext *ctx = inlink->dst; |
637 |
|
|
AudioStatsContext *s = ctx->priv; |
638 |
|
|
AVDictionary **metadata = &buf->metadata; |
639 |
|
|
|
640 |
|
|
if (s->reset_count > 0) { |
641 |
|
|
if (s->nb_frames >= s->reset_count) { |
642 |
|
|
reset_stats(s); |
643 |
|
|
s->nb_frames = 0; |
644 |
|
|
} |
645 |
|
|
s->nb_frames++; |
646 |
|
|
} |
647 |
|
|
|
648 |
|
|
ctx->internal->execute(ctx, filter_channel, buf, NULL, FFMIN(inlink->channels, ff_filter_get_nb_threads(ctx))); |
649 |
|
|
|
650 |
|
|
if (s->metadata) |
651 |
|
|
set_metadata(s, metadata); |
652 |
|
|
|
653 |
|
|
return ff_filter_frame(inlink->dst->outputs[0], buf); |
654 |
|
|
} |
655 |
|
|
|
656 |
|
|
static void print_stats(AVFilterContext *ctx) |
657 |
|
|
{ |
658 |
|
|
AudioStatsContext *s = ctx->priv; |
659 |
|
|
uint64_t mask = 0, imask = 0xFFFFFFFFFFFFFFFF, min_count = 0, max_count = 0, nb_samples = 0, noise_floor_count = 0; |
660 |
|
|
uint64_t nb_nans = 0, nb_infs = 0, nb_denormals = 0; |
661 |
|
|
double min_runs = 0, max_runs = 0, |
662 |
|
|
min = DBL_MAX, max =-DBL_MAX, min_diff = DBL_MAX, max_diff = 0, |
663 |
|
|
nmin = DBL_MAX, nmax =-DBL_MAX, |
664 |
|
|
max_sigma_x = 0, |
665 |
|
|
diff1_sum_x2 = 0, |
666 |
|
|
diff1_sum = 0, |
667 |
|
|
sigma_x = 0, |
668 |
|
|
sigma_x2 = 0, |
669 |
|
|
noise_floor = 0, |
670 |
|
|
min_sigma_x2 = DBL_MAX, |
671 |
|
|
max_sigma_x2 =-DBL_MAX; |
672 |
|
|
AVRational depth; |
673 |
|
|
int c; |
674 |
|
|
|
675 |
|
|
for (c = 0; c < s->nb_channels; c++) { |
676 |
|
|
ChannelStats *p = &s->chstats[c]; |
677 |
|
|
|
678 |
|
|
if (p->nb_samples < s->tc_samples) |
679 |
|
|
p->min_sigma_x2 = p->max_sigma_x2 = p->sigma_x2 / p->nb_samples; |
680 |
|
|
|
681 |
|
|
min = FFMIN(min, p->min); |
682 |
|
|
max = FFMAX(max, p->max); |
683 |
|
|
nmin = FFMIN(nmin, p->nmin); |
684 |
|
|
nmax = FFMAX(nmax, p->nmax); |
685 |
|
|
min_diff = FFMIN(min_diff, p->min_diff); |
686 |
|
|
max_diff = FFMAX(max_diff, p->max_diff); |
687 |
|
|
diff1_sum_x2 += p->diff1_sum_x2; |
688 |
|
|
diff1_sum += p->diff1_sum; |
689 |
|
|
min_sigma_x2 = FFMIN(min_sigma_x2, p->min_sigma_x2); |
690 |
|
|
max_sigma_x2 = FFMAX(max_sigma_x2, p->max_sigma_x2); |
691 |
|
|
sigma_x += p->sigma_x; |
692 |
|
|
sigma_x2 += p->sigma_x2; |
693 |
|
|
noise_floor = FFMAX(noise_floor, p->noise_floor); |
694 |
|
|
min_count += p->min_count; |
695 |
|
|
max_count += p->max_count; |
696 |
|
|
noise_floor_count += p->noise_floor_count; |
697 |
|
|
min_runs += p->min_runs; |
698 |
|
|
max_runs += p->max_runs; |
699 |
|
|
mask |= p->mask; |
700 |
|
|
imask &= p->imask; |
701 |
|
|
nb_samples += p->nb_samples; |
702 |
|
|
nb_nans += p->nb_nans; |
703 |
|
|
nb_infs += p->nb_infs; |
704 |
|
|
nb_denormals += p->nb_denormals; |
705 |
|
|
if (fabs(p->sigma_x) > fabs(max_sigma_x)) |
706 |
|
|
max_sigma_x = p->sigma_x; |
707 |
|
|
|
708 |
|
|
av_log(ctx, AV_LOG_INFO, "Channel: %d\n", c + 1); |
709 |
|
|
if (s->measure_perchannel & MEASURE_DC_OFFSET) |
710 |
|
|
av_log(ctx, AV_LOG_INFO, "DC offset: %f\n", p->sigma_x / p->nb_samples); |
711 |
|
|
if (s->measure_perchannel & MEASURE_MIN_LEVEL) |
712 |
|
|
av_log(ctx, AV_LOG_INFO, "Min level: %f\n", p->min); |
713 |
|
|
if (s->measure_perchannel & MEASURE_MAX_LEVEL) |
714 |
|
|
av_log(ctx, AV_LOG_INFO, "Max level: %f\n", p->max); |
715 |
|
|
if (s->measure_perchannel & MEASURE_MIN_DIFFERENCE) |
716 |
|
|
av_log(ctx, AV_LOG_INFO, "Min difference: %f\n", p->min_diff); |
717 |
|
|
if (s->measure_perchannel & MEASURE_MAX_DIFFERENCE) |
718 |
|
|
av_log(ctx, AV_LOG_INFO, "Max difference: %f\n", p->max_diff); |
719 |
|
|
if (s->measure_perchannel & MEASURE_MEAN_DIFFERENCE) |
720 |
|
|
av_log(ctx, AV_LOG_INFO, "Mean difference: %f\n", p->diff1_sum / (p->nb_samples - 1)); |
721 |
|
|
if (s->measure_perchannel & MEASURE_RMS_DIFFERENCE) |
722 |
|
|
av_log(ctx, AV_LOG_INFO, "RMS difference: %f\n", sqrt(p->diff1_sum_x2 / (p->nb_samples - 1))); |
723 |
|
|
if (s->measure_perchannel & MEASURE_PEAK_LEVEL) |
724 |
|
|
av_log(ctx, AV_LOG_INFO, "Peak level dB: %f\n", LINEAR_TO_DB(FFMAX(-p->nmin, p->nmax))); |
725 |
|
|
if (s->measure_perchannel & MEASURE_RMS_LEVEL) |
726 |
|
|
av_log(ctx, AV_LOG_INFO, "RMS level dB: %f\n", LINEAR_TO_DB(sqrt(p->sigma_x2 / p->nb_samples))); |
727 |
|
|
if (s->measure_perchannel & MEASURE_RMS_PEAK) |
728 |
|
|
av_log(ctx, AV_LOG_INFO, "RMS peak dB: %f\n", LINEAR_TO_DB(sqrt(p->max_sigma_x2))); |
729 |
|
|
if (s->measure_perchannel & MEASURE_RMS_TROUGH) |
730 |
|
|
if (p->min_sigma_x2 != 1) |
731 |
|
|
av_log(ctx, AV_LOG_INFO, "RMS trough dB: %f\n",LINEAR_TO_DB(sqrt(p->min_sigma_x2))); |
732 |
|
|
if (s->measure_perchannel & MEASURE_CREST_FACTOR) |
733 |
|
|
av_log(ctx, AV_LOG_INFO, "Crest factor: %f\n", p->sigma_x2 ? FFMAX(-p->nmin, p->nmax) / sqrt(p->sigma_x2 / p->nb_samples) : 1); |
734 |
|
|
if (s->measure_perchannel & MEASURE_FLAT_FACTOR) |
735 |
|
|
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))); |
736 |
|
|
if (s->measure_perchannel & MEASURE_PEAK_COUNT) |
737 |
|
|
av_log(ctx, AV_LOG_INFO, "Peak count: %"PRId64"\n", p->min_count + p->max_count); |
738 |
|
|
if (s->measure_perchannel & MEASURE_NOISE_FLOOR) |
739 |
|
|
av_log(ctx, AV_LOG_INFO, "Noise floor dB: %f\n", LINEAR_TO_DB(p->noise_floor)); |
740 |
|
|
if (s->measure_perchannel & MEASURE_NOISE_FLOOR_COUNT) |
741 |
|
|
av_log(ctx, AV_LOG_INFO, "Noise floor count: %"PRId64"\n", p->noise_floor_count); |
742 |
|
|
if (s->measure_perchannel & MEASURE_BIT_DEPTH) { |
743 |
|
|
bit_depth(s, p->mask, p->imask, &depth); |
744 |
|
|
av_log(ctx, AV_LOG_INFO, "Bit depth: %u/%u\n", depth.num, depth.den); |
745 |
|
|
} |
746 |
|
|
if (s->measure_perchannel & MEASURE_DYNAMIC_RANGE) |
747 |
|
|
av_log(ctx, AV_LOG_INFO, "Dynamic range: %f\n", LINEAR_TO_DB(2 * FFMAX(FFABS(p->min), FFABS(p->max))/ p->min_non_zero)); |
748 |
|
|
if (s->measure_perchannel & MEASURE_ZERO_CROSSINGS) |
749 |
|
|
av_log(ctx, AV_LOG_INFO, "Zero crossings: %"PRId64"\n", p->zero_runs); |
750 |
|
|
if (s->measure_perchannel & MEASURE_ZERO_CROSSINGS_RATE) |
751 |
|
|
av_log(ctx, AV_LOG_INFO, "Zero crossings rate: %f\n", p->zero_runs/(double)p->nb_samples); |
752 |
|
|
if ((s->is_float || s->is_double) && s->measure_perchannel & MEASURE_NUMBER_OF_NANS) |
753 |
|
|
av_log(ctx, AV_LOG_INFO, "Number of NaNs: %"PRId64"\n", p->nb_nans); |
754 |
|
|
if ((s->is_float || s->is_double) && s->measure_perchannel & MEASURE_NUMBER_OF_INFS) |
755 |
|
|
av_log(ctx, AV_LOG_INFO, "Number of Infs: %"PRId64"\n", p->nb_infs); |
756 |
|
|
if ((s->is_float || s->is_double) && s->measure_perchannel & MEASURE_NUMBER_OF_DENORMALS) |
757 |
|
|
av_log(ctx, AV_LOG_INFO, "Number of denormals: %"PRId64"\n", p->nb_denormals); |
758 |
|
|
} |
759 |
|
|
|
760 |
|
|
av_log(ctx, AV_LOG_INFO, "Overall\n"); |
761 |
|
|
if (s->measure_overall & MEASURE_DC_OFFSET) |
762 |
|
|
av_log(ctx, AV_LOG_INFO, "DC offset: %f\n", max_sigma_x / (nb_samples / s->nb_channels)); |
763 |
|
|
if (s->measure_overall & MEASURE_MIN_LEVEL) |
764 |
|
|
av_log(ctx, AV_LOG_INFO, "Min level: %f\n", min); |
765 |
|
|
if (s->measure_overall & MEASURE_MAX_LEVEL) |
766 |
|
|
av_log(ctx, AV_LOG_INFO, "Max level: %f\n", max); |
767 |
|
|
if (s->measure_overall & MEASURE_MIN_DIFFERENCE) |
768 |
|
|
av_log(ctx, AV_LOG_INFO, "Min difference: %f\n", min_diff); |
769 |
|
|
if (s->measure_overall & MEASURE_MAX_DIFFERENCE) |
770 |
|
|
av_log(ctx, AV_LOG_INFO, "Max difference: %f\n", max_diff); |
771 |
|
|
if (s->measure_overall & MEASURE_MEAN_DIFFERENCE) |
772 |
|
|
av_log(ctx, AV_LOG_INFO, "Mean difference: %f\n", diff1_sum / (nb_samples - s->nb_channels)); |
773 |
|
|
if (s->measure_overall & MEASURE_RMS_DIFFERENCE) |
774 |
|
|
av_log(ctx, AV_LOG_INFO, "RMS difference: %f\n", sqrt(diff1_sum_x2 / (nb_samples - s->nb_channels))); |
775 |
|
|
if (s->measure_overall & MEASURE_PEAK_LEVEL) |
776 |
|
|
av_log(ctx, AV_LOG_INFO, "Peak level dB: %f\n", LINEAR_TO_DB(FFMAX(-nmin, nmax))); |
777 |
|
|
if (s->measure_overall & MEASURE_RMS_LEVEL) |
778 |
|
|
av_log(ctx, AV_LOG_INFO, "RMS level dB: %f\n", LINEAR_TO_DB(sqrt(sigma_x2 / nb_samples))); |
779 |
|
|
if (s->measure_overall & MEASURE_RMS_PEAK) |
780 |
|
|
av_log(ctx, AV_LOG_INFO, "RMS peak dB: %f\n", LINEAR_TO_DB(sqrt(max_sigma_x2))); |
781 |
|
|
if (s->measure_overall & MEASURE_RMS_TROUGH) |
782 |
|
|
if (min_sigma_x2 != 1) |
783 |
|
|
av_log(ctx, AV_LOG_INFO, "RMS trough dB: %f\n", LINEAR_TO_DB(sqrt(min_sigma_x2))); |
784 |
|
|
if (s->measure_overall & MEASURE_FLAT_FACTOR) |
785 |
|
|
av_log(ctx, AV_LOG_INFO, "Flat factor: %f\n", LINEAR_TO_DB((min_runs + max_runs) / (min_count + max_count))); |
786 |
|
|
if (s->measure_overall & MEASURE_PEAK_COUNT) |
787 |
|
|
av_log(ctx, AV_LOG_INFO, "Peak count: %f\n", (min_count + max_count) / (double)s->nb_channels); |
788 |
|
|
if (s->measure_overall & MEASURE_NOISE_FLOOR) |
789 |
|
|
av_log(ctx, AV_LOG_INFO, "Noise floor dB: %f\n", LINEAR_TO_DB(noise_floor)); |
790 |
|
|
if (s->measure_overall & MEASURE_NOISE_FLOOR_COUNT) |
791 |
|
|
av_log(ctx, AV_LOG_INFO, "Noise floor count: %f\n", noise_floor_count / (double)s->nb_channels); |
792 |
|
|
if (s->measure_overall & MEASURE_BIT_DEPTH) { |
793 |
|
|
bit_depth(s, mask, imask, &depth); |
794 |
|
|
av_log(ctx, AV_LOG_INFO, "Bit depth: %u/%u\n", depth.num, depth.den); |
795 |
|
|
} |
796 |
|
|
if (s->measure_overall & MEASURE_NUMBER_OF_SAMPLES) |
797 |
|
|
av_log(ctx, AV_LOG_INFO, "Number of samples: %"PRId64"\n", nb_samples / s->nb_channels); |
798 |
|
|
if ((s->is_float || s->is_double) && s->measure_overall & MEASURE_NUMBER_OF_NANS) |
799 |
|
|
av_log(ctx, AV_LOG_INFO, "Number of NaNs: %f\n", nb_nans / (float)s->nb_channels); |
800 |
|
|
if ((s->is_float || s->is_double) && s->measure_overall & MEASURE_NUMBER_OF_INFS) |
801 |
|
|
av_log(ctx, AV_LOG_INFO, "Number of Infs: %f\n", nb_infs / (float)s->nb_channels); |
802 |
|
|
if ((s->is_float || s->is_double) && s->measure_overall & MEASURE_NUMBER_OF_DENORMALS) |
803 |
|
|
av_log(ctx, AV_LOG_INFO, "Number of denormals: %f\n", nb_denormals / (float)s->nb_channels); |
804 |
|
|
} |
805 |
|
|
|
806 |
|
|
static av_cold void uninit(AVFilterContext *ctx) |
807 |
|
|
{ |
808 |
|
|
AudioStatsContext *s = ctx->priv; |
809 |
|
|
|
810 |
|
|
if (s->nb_channels) |
811 |
|
|
print_stats(ctx); |
812 |
|
|
if (s->chstats) { |
813 |
|
|
for (int i = 0; i < s->nb_channels; i++) { |
814 |
|
|
ChannelStats *p = &s->chstats[i]; |
815 |
|
|
|
816 |
|
|
av_freep(&p->win_samples); |
817 |
|
|
} |
818 |
|
|
} |
819 |
|
|
av_freep(&s->chstats); |
820 |
|
|
} |
821 |
|
|
|
822 |
|
|
static const AVFilterPad astats_inputs[] = { |
823 |
|
|
{ |
824 |
|
|
.name = "default", |
825 |
|
|
.type = AVMEDIA_TYPE_AUDIO, |
826 |
|
|
.filter_frame = filter_frame, |
827 |
|
|
}, |
828 |
|
|
{ NULL } |
829 |
|
|
}; |
830 |
|
|
|
831 |
|
|
static const AVFilterPad astats_outputs[] = { |
832 |
|
|
{ |
833 |
|
|
.name = "default", |
834 |
|
|
.type = AVMEDIA_TYPE_AUDIO, |
835 |
|
|
.config_props = config_output, |
836 |
|
|
}, |
837 |
|
|
{ NULL } |
838 |
|
|
}; |
839 |
|
|
|
840 |
|
|
AVFilter ff_af_astats = { |
841 |
|
|
.name = "astats", |
842 |
|
|
.description = NULL_IF_CONFIG_SMALL("Show time domain statistics about audio frames."), |
843 |
|
|
.query_formats = query_formats, |
844 |
|
|
.priv_size = sizeof(AudioStatsContext), |
845 |
|
|
.priv_class = &astats_class, |
846 |
|
|
.uninit = uninit, |
847 |
|
|
.inputs = astats_inputs, |
848 |
|
|
.outputs = astats_outputs, |
849 |
|
|
.flags = AVFILTER_FLAG_SLICE_THREADS, |
850 |
|
|
}; |