af_replaygain.c 24 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613
  1. /*
  2. * Copyright (c) 1998 - 2009 Conifer Software
  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
  8. * License 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 GNU
  14. * Lesser General Public License for more details.
  15. *
  16. * You should have received a copy of the GNU Lesser General Public
  17. * License along with FFmpeg; if not, write to the Free Software
  18. * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19. */
  20. /**
  21. * @file
  22. * ReplayGain scanner
  23. */
  24. #include "libavutil/avassert.h"
  25. #include "libavutil/channel_layout.h"
  26. #include "audio.h"
  27. #include "avfilter.h"
  28. #include "internal.h"
  29. #define HISTOGRAM_SLOTS 12000
  30. #define BUTTER_ORDER 2
  31. #define YULE_ORDER 10
  32. typedef struct ReplayGainFreqInfo {
  33. int sample_rate;
  34. double BYule[YULE_ORDER + 1];
  35. double AYule[YULE_ORDER + 1];
  36. double BButter[BUTTER_ORDER + 1];
  37. double AButter[BUTTER_ORDER + 1];
  38. } ReplayGainFreqInfo;
  39. static const ReplayGainFreqInfo freqinfos[] =
  40. {
  41. {
  42. 192000,
  43. { 0.01184742123123, -0.04631092400086, 0.06584226961238,
  44. -0.02165588522478, -0.05656260778952, 0.08607493592760,
  45. -0.03375544339786, -0.04216579932754, 0.06416711490648,
  46. -0.03444708260844, 0.00697275872241 },
  47. { 1.00000000000000, -5.24727318348167, 10.60821585192244,
  48. -8.74127665810413, -1.33906071371683, 8.07972882096606,
  49. -5.46179918950847, 0.54318070652536, 0.87450969224280,
  50. -0.34656083539754, 0.03034796843589 },
  51. { 0.99653501465135, -1.99307002930271, 0.99653501465135 },
  52. { 1.00000000000000, -1.99305802314321, 0.99308203546221 },
  53. },
  54. {
  55. 176400,
  56. { 0.00268568524529, -0.00852379426080, 0.00852704191347,
  57. 0.00146116310295, -0.00950855828762, 0.00625449515499,
  58. 0.00116183868722, -0.00362461417136, 0.00203961000134,
  59. -0.00050664587933, 0.00004327455427 },
  60. { 1.00000000000000, -5.57512782763045, 12.44291056065794,
  61. -12.87462799681221, 3.08554846961576, 6.62493459880692,
  62. -7.07662766313248, 2.51175542736441, 0.06731510802735,
  63. -0.24567753819213, 0.03961404162376 },
  64. { 0.99622916581118, -1.99245833162236, 0.99622916581118 },
  65. { 1.00000000000000, -1.99244411238133, 0.99247255086339 },
  66. },
  67. {
  68. 144000,
  69. { 0.00639682359450, -0.02556437970955, 0.04230854400938,
  70. -0.03722462201267, 0.01718514827295, 0.00610592243009,
  71. -0.03065965747365, 0.04345745003539, -0.03298592681309,
  72. 0.01320937236809, -0.00220304127757 },
  73. { 1.00000000000000, -6.14814623523425, 15.80002457141566,
  74. -20.78487587686937, 11.98848552310315, 3.36462015062606,
  75. -10.22419868359470, 6.65599702146473, -1.67141861110485,
  76. -0.05417956536718, 0.07374767867406 },
  77. { 0.99538268958706, -1.99076537917413, 0.99538268958706 },
  78. { 1.00000000000000, -1.99074405950505, 0.99078669884321 },
  79. },
  80. {
  81. 128000,
  82. { 0.00553120584305, -0.02112620545016, 0.03549076243117,
  83. -0.03362498312306, 0.01425867248183, 0.01344686928787,
  84. -0.03392770787836, 0.03464136459530, -0.02039116051549,
  85. 0.00667420794705, -0.00093763762995 },
  86. { 1.00000000000000, -6.14581710839925, 16.04785903675838,
  87. -22.19089131407749, 15.24756471580286, -0.52001440400238,
  88. -8.00488641699940, 6.60916094768855, -2.37856022810923,
  89. 0.33106947986101, 0.00459820832036 },
  90. { 0.99480702681278, -1.98961405362557, 0.99480702681278 },
  91. { 1.00000000000000, -1.98958708647324, 0.98964102077790 },
  92. },
  93. {
  94. 112000,
  95. { 0.00528778718259, -0.01893240907245, 0.03185982561867,
  96. -0.02926260297838, 0.00715743034072, 0.01985743355827,
  97. -0.03222614850941, 0.02565681978192, -0.01210662313473,
  98. 0.00325436284541, -0.00044173593001 },
  99. { 1.00000000000000, -6.24932108456288, 17.42344320538476,
  100. -27.86819709054896, 26.79087344681326,-13.43711081485123,
  101. -0.66023612948173, 6.03658091814935, -4.24926577030310,
  102. 1.40829268709186, -0.19480852628112 },
  103. { 0.99406737810867, -1.98813475621734, 0.99406737810867 },
  104. { 1.00000000000000, -1.98809955990514, 0.98816995252954 },
  105. },
  106. {
  107. 96000,
  108. { 0.00588138296683, -0.01613559730421, 0.02184798954216,
  109. -0.01742490405317, 0.00464635643780, 0.01117772513205,
  110. -0.02123865824368, 0.01959354413350, -0.01079720643523,
  111. 0.00352183686289, -0.00063124341421 },
  112. { 1.00000000000000, -5.97808823642008, 16.21362507964068,
  113. -25.72923730652599, 25.40470663139513,-14.66166287771134,
  114. 2.81597484359752, 2.51447125969733, -2.23575306985286,
  115. 0.75788151036791, -0.10078025199029 },
  116. { 0.99308203517541, -1.98616407035082, 0.99308203517541 },
  117. { 1.00000000000000, -1.98611621154089, 0.98621192916075 },
  118. },
  119. {
  120. 88200,
  121. { 0.02667482047416, -0.11377479336097, 0.23063167910965,
  122. -0.30726477945593, 0.33188520686529, -0.33862680249063,
  123. 0.31807161531340, -0.23730796929880, 0.12273894790371,
  124. -0.03840017967282, 0.00549673387936 },
  125. { 1.00000000000000, -6.31836451657302, 18.31351310801799,
  126. -31.88210014815921, 36.53792146976740,-28.23393036467559,
  127. 14.24725258227189, -4.04670980012854, 0.18865757280515,
  128. 0.25420333563908, -0.06012333531065 },
  129. { 0.99247255046129, -1.98494510092259, 0.99247255046129 },
  130. { 1.00000000000000, -1.98488843762335, 0.98500176422183 },
  131. },
  132. {
  133. 64000,
  134. { 0.02613056568174, -0.08128786488109, 0.14937282347325,
  135. -0.21695711675126, 0.25010286673402, -0.23162283619278,
  136. 0.17424041833052, -0.10299599216680, 0.04258696481981,
  137. -0.00977952936493, 0.00105325558889 },
  138. { 1.00000000000000, -5.73625477092119, 16.15249794355035,
  139. -29.68654912464508, 39.55706155674083,-39.82524556246253,
  140. 30.50605345013009,-17.43051772821245, 7.05154573908017,
  141. -1.80783839720514, 0.22127840210813 },
  142. { 0.98964101933472, -1.97928203866944, 0.98964101933472 },
  143. { 1.00000000000000, -1.97917472731009, 0.97938935002880 },
  144. },
  145. {
  146. 56000,
  147. { 0.03144914734085, -0.06151729206963, 0.08066788708145,
  148. -0.09737939921516, 0.08943210803999, -0.06989984672010,
  149. 0.04926972841044, -0.03161257848451, 0.01456837493506,
  150. -0.00316015108496, 0.00132807215875 },
  151. { 1.00000000000000, -4.87377313090032, 12.03922160140209,
  152. -20.10151118381395, 25.10388534415171,-24.29065560815903,
  153. 18.27158469090663,-10.45249552560593, 4.30319491872003,
  154. -1.13716992070185, 0.14510733527035 },
  155. { 0.98816995007392, -1.97633990014784, 0.98816995007392 },
  156. { 1.00000000000000, -1.97619994516973, 0.97647985512594 },
  157. },
  158. {
  159. 48000,
  160. { 0.03857599435200, -0.02160367184185, -0.00123395316851,
  161. -0.00009291677959, -0.01655260341619, 0.02161526843274,
  162. -0.02074045215285, 0.00594298065125, 0.00306428023191,
  163. 0.00012025322027, 0.00288463683916 },
  164. { 1.00000000000000, -3.84664617118067, 7.81501653005538,
  165. -11.34170355132042, 13.05504219327545,-12.28759895145294,
  166. 9.48293806319790, -5.87257861775999, 2.75465861874613,
  167. -0.86984376593551, 0.13919314567432 },
  168. { 0.98621192462708, -1.97242384925416, 0.98621192462708 },
  169. { 1.00000000000000, -1.97223372919527, 0.97261396931306 },
  170. },
  171. {
  172. 44100,
  173. { 0.05418656406430, -0.02911007808948, -0.00848709379851,
  174. -0.00851165645469, -0.00834990904936, 0.02245293253339,
  175. -0.02596338512915, 0.01624864962975, -0.00240879051584,
  176. 0.00674613682247, -0.00187763777362 },
  177. { 1.00000000000000, -3.47845948550071, 6.36317777566148,
  178. -8.54751527471874, 9.47693607801280, -8.81498681370155,
  179. 6.85401540936998, -4.39470996079559, 2.19611684890774,
  180. -0.75104302451432, 0.13149317958808 },
  181. { 0.98500175787242, -1.97000351574484, 0.98500175787242 },
  182. { 1.00000000000000, -1.96977855582618, 0.97022847566350 },
  183. },
  184. {
  185. 37800,
  186. { 0.08717879977844, -0.01000374016172, -0.06265852122368,
  187. -0.01119328800950, -0.00114279372960, 0.02081333954769,
  188. -0.01603261863207, 0.01936763028546, 0.00760044736442,
  189. -0.00303979112271, -0.00075088605788 },
  190. { 1.00000000000000, -2.62816311472146, 3.53734535817992,
  191. -3.81003448678921, 3.91291636730132, -3.53518605896288,
  192. 2.71356866157873, -1.86723311846592, 1.12075382367659,
  193. -0.48574086886890, 0.11330544663849 },
  194. { 0.98252400815195, -1.96504801630391, 0.98252400815195 },
  195. { 1.00000000000000, -1.96474258269041, 0.96535344991740 },
  196. },
  197. {
  198. 32000,
  199. { 0.15457299681924, -0.09331049056315, -0.06247880153653,
  200. 0.02163541888798, -0.05588393329856, 0.04781476674921,
  201. 0.00222312597743, 0.03174092540049, -0.01390589421898,
  202. 0.00651420667831, -0.00881362733839 },
  203. { 1.00000000000000, -2.37898834973084, 2.84868151156327,
  204. -2.64577170229825, 2.23697657451713, -1.67148153367602,
  205. 1.00595954808547, -0.45953458054983, 0.16378164858596,
  206. -0.05032077717131, 0.02347897407020 },
  207. { 0.97938932735214, -1.95877865470428, 0.97938932735214 },
  208. { 1.00000000000000, -1.95835380975398, 0.95920349965459 },
  209. },
  210. {
  211. 24000,
  212. { 0.30296907319327, -0.22613988682123, -0.08587323730772,
  213. 0.03282930172664, -0.00915702933434, -0.02364141202522,
  214. -0.00584456039913, 0.06276101321749, -0.00000828086748,
  215. 0.00205861885564, -0.02950134983287 },
  216. { 1.00000000000000, -1.61273165137247, 1.07977492259970,
  217. -0.25656257754070, -0.16276719120440, -0.22638893773906,
  218. 0.39120800788284, -0.22138138954925, 0.04500235387352,
  219. 0.02005851806501, 0.00302439095741 },
  220. { 0.97531843204928, -1.95063686409857, 0.97531843204928 },
  221. { 1.00000000000000, -1.95002759149878, 0.95124613669835 },
  222. },
  223. {
  224. 22050,
  225. { 0.33642304856132, -0.25572241425570, -0.11828570177555,
  226. 0.11921148675203, -0.07834489609479, -0.00469977914380,
  227. -0.00589500224440, 0.05724228140351, 0.00832043980773,
  228. -0.01635381384540, -0.01760176568150 },
  229. { 1.00000000000000, -1.49858979367799, 0.87350271418188,
  230. 0.12205022308084, -0.80774944671438, 0.47854794562326,
  231. -0.12453458140019, -0.04067510197014, 0.08333755284107,
  232. -0.04237348025746, 0.02977207319925 },
  233. { 0.97316523498161, -1.94633046996323, 0.97316523498161 },
  234. { 1.00000000000000, -1.94561023566527, 0.94705070426118 },
  235. },
  236. {
  237. 18900,
  238. { 0.38524531015142, -0.27682212062067, -0.09980181488805,
  239. 0.09951486755646, -0.08934020156622, -0.00322369330199,
  240. -0.00110329090689, 0.03784509844682, 0.01683906213303,
  241. -0.01147039862572, -0.01941767987192 },
  242. { 1.00000000000000, -1.29708918404534, 0.90399339674203,
  243. -0.29613799017877, -0.42326645916207, 0.37934887402200,
  244. -0.37919795944938, 0.23410283284785, -0.03892971758879,
  245. 0.00403009552351, 0.03640166626278 },
  246. { 0.96535326815829, -1.93070653631658, 0.96535326815829 },
  247. { 1.00000000000000, -1.92950577983524, 0.93190729279793 },
  248. },
  249. {
  250. 16000,
  251. { 0.44915256608450, -0.14351757464547, -0.22784394429749,
  252. -0.01419140100551, 0.04078262797139, -0.12398163381748,
  253. 0.04097565135648, 0.10478503600251, -0.01863887810927,
  254. -0.03193428438915, 0.00541907748707 },
  255. { 1.00000000000000, -0.62820619233671, 0.29661783706366,
  256. -0.37256372942400, 0.00213767857124, -0.42029820170918,
  257. 0.22199650564824, 0.00613424350682, 0.06747620744683,
  258. 0.05784820375801, 0.03222754072173 },
  259. { 0.96454515552826, -1.92909031105652, 0.96454515552826 },
  260. { 1.00000000000000, -1.92783286977036, 0.93034775234268 },
  261. },
  262. {
  263. 12000,
  264. { 0.56619470757641, -0.75464456939302, 0.16242137742230,
  265. 0.16744243493672, -0.18901604199609, 0.30931782841830,
  266. -0.27562961986224, 0.00647310677246, 0.08647503780351,
  267. -0.03788984554840, -0.00588215443421 },
  268. { 1.00000000000000, -1.04800335126349, 0.29156311971249,
  269. -0.26806001042947, 0.00819999645858, 0.45054734505008,
  270. -0.33032403314006, 0.06739368333110, -0.04784254229033,
  271. 0.01639907836189, 0.01807364323573 },
  272. { 0.96009142950541, -1.92018285901082, 0.96009142950541 },
  273. { 1.00000000000000, -1.91858953033784, 0.92177618768381 },
  274. },
  275. {
  276. 11025,
  277. { 0.58100494960553, -0.53174909058578, -0.14289799034253,
  278. 0.17520704835522, 0.02377945217615, 0.15558449135573,
  279. -0.25344790059353, 0.01628462406333, 0.06920467763959,
  280. -0.03721611395801, -0.00749618797172 },
  281. { 1.00000000000000, -0.51035327095184, -0.31863563325245,
  282. -0.20256413484477, 0.14728154134330, 0.38952639978999,
  283. -0.23313271880868, -0.05246019024463, -0.02505961724053,
  284. 0.02442357316099, 0.01818801111503 },
  285. { 0.95856916599601, -1.91713833199203, 0.95856916599601 },
  286. { 1.00000000000000, -1.91542108074780, 0.91885558323625 },
  287. },
  288. {
  289. 8000,
  290. { 0.53648789255105, -0.42163034350696, -0.00275953611929,
  291. 0.04267842219415, -0.10214864179676, 0.14590772289388,
  292. -0.02459864859345, -0.11202315195388, -0.04060034127000,
  293. 0.04788665548180, -0.02217936801134 },
  294. { 1.00000000000000, -0.25049871956020, -0.43193942311114,
  295. -0.03424681017675, -0.04678328784242, 0.26408300200955,
  296. 0.15113130533216, -0.17556493366449, -0.18823009262115,
  297. 0.05477720428674, 0.04704409688120 },
  298. { 0.94597685600279, -1.89195371200558, 0.94597685600279 },
  299. { 1.00000000000000, -1.88903307939452, 0.89487434461664 },
  300. },
  301. };
  302. typedef struct ReplayGainContext {
  303. uint32_t histogram[HISTOGRAM_SLOTS];
  304. float peak;
  305. int yule_hist_i, butter_hist_i;
  306. const double *yule_coeff_a;
  307. const double *yule_coeff_b;
  308. const double *butter_coeff_a;
  309. const double *butter_coeff_b;
  310. float yule_hist_a[256];
  311. float yule_hist_b[256];
  312. float butter_hist_a[256];
  313. float butter_hist_b[256];
  314. } ReplayGainContext;
  315. static int query_formats(AVFilterContext *ctx)
  316. {
  317. AVFilterFormats *formats = NULL;
  318. AVFilterChannelLayouts *layout = NULL;
  319. int i;
  320. ff_add_format(&formats, AV_SAMPLE_FMT_FLT);
  321. ff_set_common_formats(ctx, formats);
  322. ff_add_channel_layout(&layout, AV_CH_LAYOUT_STEREO);
  323. ff_set_common_channel_layouts(ctx, layout);
  324. formats = NULL;
  325. for (i = 0; i < FF_ARRAY_ELEMS(freqinfos); i++)
  326. ff_add_format(&formats, freqinfos[i].sample_rate);
  327. ff_set_common_samplerates(ctx, formats);
  328. return 0;
  329. }
  330. static int config_input(AVFilterLink *inlink)
  331. {
  332. AVFilterContext *ctx = inlink->dst;
  333. ReplayGainContext *s = ctx->priv;
  334. int i;
  335. for (i = 0; i < FF_ARRAY_ELEMS(freqinfos); i++) {
  336. if (freqinfos[i].sample_rate == inlink->sample_rate)
  337. break;
  338. }
  339. av_assert0(i < FF_ARRAY_ELEMS(freqinfos));
  340. s->yule_coeff_a = freqinfos[i].AYule;
  341. s->yule_coeff_b = freqinfos[i].BYule;
  342. s->butter_coeff_a = freqinfos[i].AButter;
  343. s->butter_coeff_b = freqinfos[i].BButter;
  344. s->yule_hist_i = 20;
  345. s->butter_hist_i = 4;
  346. inlink->partial_buf_size =
  347. inlink->min_samples =
  348. inlink->max_samples = inlink->sample_rate / 20;
  349. return 0;
  350. }
  351. /*
  352. * Update largest absolute sample value.
  353. */
  354. static void calc_stereo_peak(const float *samples, int nb_samples,
  355. float *peak_p)
  356. {
  357. float peak = 0.0;
  358. while (nb_samples--) {
  359. if (samples[0] > peak)
  360. peak = samples[0];
  361. else if (-samples[0] > peak)
  362. peak = -samples[0];
  363. if (samples[1] > peak)
  364. peak = samples[1];
  365. else if (-samples[1] > peak)
  366. peak = -samples[1];
  367. samples += 2;
  368. }
  369. *peak_p = FFMAX(peak, *peak_p);
  370. }
  371. /*
  372. * Calculate stereo RMS level. Minimum value is about -100 dB for
  373. * digital silence. The 90 dB offset is to compensate for the
  374. * normalized float range and 3 dB is for stereo samples.
  375. */
  376. static double calc_stereo_rms(const float *samples, int nb_samples)
  377. {
  378. int count = nb_samples;
  379. double sum = 1e-16;
  380. while (count--) {
  381. sum += samples[0] * samples[0] + samples[1] * samples[1];
  382. samples += 2;
  383. }
  384. return 10 * log10 (sum / nb_samples) + 90.0 - 3.0;
  385. }
  386. /*
  387. * Optimized implementation of 2nd-order IIR stereo filter.
  388. */
  389. static void butter_filter_stereo_samples(ReplayGainContext *s,
  390. float *samples, int nb_samples)
  391. {
  392. const double *coeff_a = s->butter_coeff_a;
  393. const double *coeff_b = s->butter_coeff_b;
  394. float *hist_a = s->butter_hist_a;
  395. float *hist_b = s->butter_hist_b;
  396. double left, right;
  397. int i, j;
  398. i = s->butter_hist_i;
  399. // If filter history is very small magnitude, clear it completely
  400. // to prevent denormals from rattling around in there forever
  401. // (slowing us down).
  402. for (j = -4; j < 0; ++j)
  403. if (fabs(hist_a[i + j]) > 1e-10 || fabs(hist_b[i + j]) > 1e-10)
  404. break;
  405. if (!j) {
  406. memset(s->butter_hist_a, 0, sizeof(s->butter_hist_a));
  407. memset(s->butter_hist_b, 0, sizeof(s->butter_hist_b));
  408. }
  409. while (nb_samples--) {
  410. left = (hist_b[i ] = samples[0]) * coeff_b[0];
  411. right = (hist_b[i + 1] = samples[1]) * coeff_b[0];
  412. left += hist_b[i - 2] * coeff_b[1] - hist_a[i - 2] * coeff_a[1];
  413. right += hist_b[i - 1] * coeff_b[1] - hist_a[i - 1] * coeff_a[1];
  414. left += hist_b[i - 4] * coeff_b[2] - hist_a[i - 4] * coeff_a[2];
  415. right += hist_b[i - 3] * coeff_b[2] - hist_a[i - 3] * coeff_a[2];
  416. samples[0] = hist_a[i ] = (float) left;
  417. samples[1] = hist_a[i + 1] = (float) right;
  418. samples += 2;
  419. if ((i += 2) == 256) {
  420. memcpy(hist_a, hist_a + 252, sizeof(*hist_a) * 4);
  421. memcpy(hist_b, hist_b + 252, sizeof(*hist_b) * 4);
  422. i = 4;
  423. }
  424. }
  425. s->butter_hist_i = i;
  426. }
  427. /*
  428. * Optimized implementation of 10th-order IIR stereo filter.
  429. */
  430. static void yule_filter_stereo_samples(ReplayGainContext *s, const float *src,
  431. float *dst, int nb_samples)
  432. {
  433. const double *coeff_a = s->yule_coeff_a;
  434. const double *coeff_b = s->yule_coeff_b;
  435. float *hist_a = s->yule_hist_a;
  436. float *hist_b = s->yule_hist_b;
  437. double left, right;
  438. int i, j;
  439. i = s->yule_hist_i;
  440. // If filter history is very small magnitude, clear it completely to
  441. // prevent denormals from rattling around in there forever
  442. // (slowing us down).
  443. for (j = -20; j < 0; ++j)
  444. if (fabs(hist_a[i + j]) > 1e-10 || fabs(hist_b[i + j]) > 1e-10)
  445. break;
  446. if (!j) {
  447. memset(s->yule_hist_a, 0, sizeof(s->yule_hist_a));
  448. memset(s->yule_hist_b, 0, sizeof(s->yule_hist_b));
  449. }
  450. while (nb_samples--) {
  451. left = (hist_b[i] = src[0]) * coeff_b[0];
  452. right = (hist_b[i + 1] = src[1]) * coeff_b[0];
  453. left += hist_b[i - 2] * coeff_b[ 1] - hist_a[i - 2] * coeff_a[1 ];
  454. right += hist_b[i - 1] * coeff_b[ 1] - hist_a[i - 1] * coeff_a[1 ];
  455. left += hist_b[i - 4] * coeff_b[ 2] - hist_a[i - 4] * coeff_a[2 ];
  456. right += hist_b[i - 3] * coeff_b[ 2] - hist_a[i - 3] * coeff_a[2 ];
  457. left += hist_b[i - 6] * coeff_b[ 3] - hist_a[i - 6] * coeff_a[3 ];
  458. right += hist_b[i - 5] * coeff_b[ 3] - hist_a[i - 5] * coeff_a[3 ];
  459. left += hist_b[i - 8] * coeff_b[ 4] - hist_a[i - 8] * coeff_a[4 ];
  460. right += hist_b[i - 7] * coeff_b[ 4] - hist_a[i - 7] * coeff_a[4 ];
  461. left += hist_b[i - 10] * coeff_b[ 5] - hist_a[i - 10] * coeff_a[5 ];
  462. right += hist_b[i - 9] * coeff_b[ 5] - hist_a[i - 9] * coeff_a[5 ];
  463. left += hist_b[i - 12] * coeff_b[ 6] - hist_a[i - 12] * coeff_a[6 ];
  464. right += hist_b[i - 11] * coeff_b[ 6] - hist_a[i - 11] * coeff_a[6 ];
  465. left += hist_b[i - 14] * coeff_b[ 7] - hist_a[i - 14] * coeff_a[7 ];
  466. right += hist_b[i - 13] * coeff_b[ 7] - hist_a[i - 13] * coeff_a[7 ];
  467. left += hist_b[i - 16] * coeff_b[ 8] - hist_a[i - 16] * coeff_a[8 ];
  468. right += hist_b[i - 15] * coeff_b[ 8] - hist_a[i - 15] * coeff_a[8 ];
  469. left += hist_b[i - 18] * coeff_b[ 9] - hist_a[i - 18] * coeff_a[9 ];
  470. right += hist_b[i - 17] * coeff_b[ 9] - hist_a[i - 17] * coeff_a[9 ];
  471. left += hist_b[i - 20] * coeff_b[10] - hist_a[i - 20] * coeff_a[10];
  472. right += hist_b[i - 19] * coeff_b[10] - hist_a[i - 19] * coeff_a[10];
  473. dst[0] = hist_a[i ] = (float)left;
  474. dst[1] = hist_a[i + 1] = (float)right;
  475. src += 2;
  476. dst += 2;
  477. if ((i += 2) == 256) {
  478. memcpy(hist_a, hist_a + 236, sizeof(*hist_a) * 20);
  479. memcpy(hist_b, hist_b + 236, sizeof(*hist_b) * 20);
  480. i = 20;
  481. }
  482. }
  483. s->yule_hist_i = i;
  484. }
  485. /*
  486. * Calculate the ReplayGain value from the specified loudness histogram;
  487. * clip to -24 / +64 dB.
  488. */
  489. static float calc_replaygain(uint32_t *histogram)
  490. {
  491. uint32_t loud_count = 0, total_windows = 0;
  492. float gain;
  493. int i;
  494. for (i = 0; i < HISTOGRAM_SLOTS; i++)
  495. total_windows += histogram [i];
  496. while (i--)
  497. if ((loud_count += histogram [i]) * 20 >= total_windows)
  498. break;
  499. gain = (float)(64.54 - i / 100.0);
  500. return av_clipf(gain, -24.0, 64.0);
  501. }
  502. static int filter_frame(AVFilterLink *inlink, AVFrame *in)
  503. {
  504. AVFilterContext *ctx = inlink->dst;
  505. AVFilterLink *outlink = ctx->outputs[0];
  506. ReplayGainContext *s = ctx->priv;
  507. uint32_t level;
  508. AVFrame *out;
  509. out = ff_get_audio_buffer(inlink, in->nb_samples);
  510. if (!out) {
  511. av_frame_free(&in);
  512. return AVERROR(ENOMEM);
  513. }
  514. calc_stereo_peak((float *)in->data[0],
  515. in->nb_samples, &s->peak);
  516. yule_filter_stereo_samples(s, (const float *)in->data[0],
  517. (float *)out->data[0],
  518. out->nb_samples);
  519. butter_filter_stereo_samples(s, (float *)out->data[0],
  520. out->nb_samples);
  521. level = (uint32_t)floor(100 * calc_stereo_rms((float *)out->data[0],
  522. out->nb_samples));
  523. level = av_clip(level, 0, HISTOGRAM_SLOTS - 1);
  524. s->histogram[level]++;
  525. av_frame_free(&out);
  526. return ff_filter_frame(outlink, in);
  527. }
  528. static av_cold void uninit(AVFilterContext *ctx)
  529. {
  530. ReplayGainContext *s = ctx->priv;
  531. float gain = calc_replaygain(s->histogram);
  532. av_log(ctx, AV_LOG_INFO, "track_gain = %+.2f dB\n", gain);
  533. av_log(ctx, AV_LOG_INFO, "track_peak = %.6f\n", s->peak);
  534. }
  535. static const AVFilterPad replaygain_inputs[] = {
  536. {
  537. .name = "default",
  538. .type = AVMEDIA_TYPE_AUDIO,
  539. .filter_frame = filter_frame,
  540. .config_props = config_input,
  541. },
  542. { NULL }
  543. };
  544. static const AVFilterPad replaygain_outputs[] = {
  545. {
  546. .name = "default",
  547. .type = AVMEDIA_TYPE_AUDIO,
  548. },
  549. { NULL }
  550. };
  551. AVFilter ff_af_replaygain = {
  552. .name = "replaygain",
  553. .description = NULL_IF_CONFIG_SMALL("ReplayGain scanner."),
  554. .query_formats = query_formats,
  555. .uninit = uninit,
  556. .priv_size = sizeof(ReplayGainContext),
  557. .inputs = replaygain_inputs,
  558. .outputs = replaygain_outputs,
  559. };