1 /*
2 * Copyright (c) 2016 Paul B Mahol
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU Lesser General Public License as published
8 * by the Free Software Foundation; either version 2.1 of the License,
9 * 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 #include "libavutil/avstring.h"
22 #include "libavfilter/internal.h"
23 #include "libavutil/common.h"
24 #include "libavutil/cpu.h"
25 #include "libavutil/opt.h"
26 #include "libavutil/eval.h"
27 #include "libavutil/tx.h"
28 #include "audio.h"
29 #include "filters.h"
30 #include "window_func.h"
31
32 typedef struct AFFTFiltContext {
33 const AVClass *class;
34 char *real_str;
35 char *img_str;
36 int fft_size;
37
38 AVTXContext **fft, **ifft;
39 av_tx_fn tx_fn, itx_fn;
40 AVComplexFloat **fft_in;
41 AVComplexFloat **fft_out;
42 AVComplexFloat **fft_temp;
43 int nb_exprs;
44 int channels;
45 int window_size;
46 AVExpr **real;
47 AVExpr **imag;
48 int hop_size;
49 float overlap;
50 AVFrame *window;
51 AVFrame *buffer;
52 int win_func;
53 float *window_func_lut;
54 } AFFTFiltContext;
55
56 static const char *const var_names[] = { "sr", "b", "nb", "ch", "chs", "pts", "re", "im", NULL };
57 enum { VAR_SAMPLE_RATE, VAR_BIN, VAR_NBBINS, VAR_CHANNEL, VAR_CHANNELS, VAR_PTS, VAR_REAL, VAR_IMAG, VAR_VARS_NB };
58
59 #define OFFSET(x) offsetof(AFFTFiltContext, x)
60 #define A AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
61
62 static const AVOption afftfilt_options[] = {
63 { "real", "set channels real expressions", OFFSET(real_str), AV_OPT_TYPE_STRING, {.str = "re" }, 0, 0, A },
64 { "imag", "set channels imaginary expressions", OFFSET(img_str), AV_OPT_TYPE_STRING, {.str = "im" }, 0, 0, A },
65 { "win_size", "set window size", OFFSET(fft_size), AV_OPT_TYPE_INT, {.i64=4096}, 16, 131072, A },
66 WIN_FUNC_OPTION("win_func", OFFSET(win_func), A, WFUNC_HANNING),
67 { "overlap", "set window overlap", OFFSET(overlap), AV_OPT_TYPE_FLOAT, {.dbl=0.75}, 0, 1, A },
68 { NULL },
69 };
70
71 AVFILTER_DEFINE_CLASS(afftfilt);
72
getreal(void * priv,double x,double ch)73 static inline double getreal(void *priv, double x, double ch)
74 {
75 AFFTFiltContext *s = priv;
76 int ich, ix;
77
78 ich = av_clip(ch, 0, s->nb_exprs - 1);
79 ix = av_clip(x, 0, s->window_size / 2);
80
81 return s->fft_out[ich][ix].re;
82 }
83
getimag(void * priv,double x,double ch)84 static inline double getimag(void *priv, double x, double ch)
85 {
86 AFFTFiltContext *s = priv;
87 int ich, ix;
88
89 ich = av_clip(ch, 0, s->nb_exprs - 1);
90 ix = av_clip(x, 0, s->window_size / 2);
91
92 return s->fft_out[ich][ix].im;
93 }
94
realf(void * priv,double x,double ch)95 static double realf(void *priv, double x, double ch) { return getreal(priv, x, ch); }
imagf(void * priv,double x,double ch)96 static double imagf(void *priv, double x, double ch) { return getimag(priv, x, ch); }
97
98 static const char *const func2_names[] = { "real", "imag", NULL };
99 static double (*const func2[])(void *, double, double) = { realf, imagf, NULL };
100
config_input(AVFilterLink * inlink)101 static int config_input(AVFilterLink *inlink)
102 {
103 AVFilterContext *ctx = inlink->dst;
104 AFFTFiltContext *s = ctx->priv;
105 char *saveptr = NULL;
106 int ret = 0, ch;
107 float overlap, scale = 1.f;
108 char *args;
109 const char *last_expr = "1";
110 int buf_size;
111
112 s->channels = inlink->ch_layout.nb_channels;
113 s->fft = av_calloc(s->channels, sizeof(*s->fft));
114 s->ifft = av_calloc(s->channels, sizeof(*s->ifft));
115 if (!s->fft || !s->ifft)
116 return AVERROR(ENOMEM);
117
118 for (int ch = 0; ch < s->channels; ch++) {
119 ret = av_tx_init(&s->fft[ch], &s->tx_fn, AV_TX_FLOAT_FFT, 0, s->fft_size, &scale, 0);
120 if (ret < 0)
121 return ret;
122 }
123
124 for (int ch = 0; ch < s->channels; ch++) {
125 ret = av_tx_init(&s->ifft[ch], &s->itx_fn, AV_TX_FLOAT_FFT, 1, s->fft_size, &scale, 0);
126 if (ret < 0)
127 return ret;
128 }
129
130 s->window_size = s->fft_size;
131 buf_size = FFALIGN(s->window_size, av_cpu_max_align());
132
133 s->fft_in = av_calloc(inlink->ch_layout.nb_channels, sizeof(*s->fft_in));
134 if (!s->fft_in)
135 return AVERROR(ENOMEM);
136
137 s->fft_out = av_calloc(inlink->ch_layout.nb_channels, sizeof(*s->fft_out));
138 if (!s->fft_out)
139 return AVERROR(ENOMEM);
140
141 s->fft_temp = av_calloc(inlink->ch_layout.nb_channels, sizeof(*s->fft_temp));
142 if (!s->fft_temp)
143 return AVERROR(ENOMEM);
144
145 for (ch = 0; ch < inlink->ch_layout.nb_channels; ch++) {
146 s->fft_in[ch] = av_calloc(buf_size, sizeof(**s->fft_in));
147 if (!s->fft_in[ch])
148 return AVERROR(ENOMEM);
149
150 s->fft_out[ch] = av_calloc(buf_size, sizeof(**s->fft_out));
151 if (!s->fft_out[ch])
152 return AVERROR(ENOMEM);
153
154 s->fft_temp[ch] = av_calloc(buf_size, sizeof(**s->fft_temp));
155 if (!s->fft_temp[ch])
156 return AVERROR(ENOMEM);
157 }
158
159 s->real = av_calloc(inlink->ch_layout.nb_channels, sizeof(*s->real));
160 if (!s->real)
161 return AVERROR(ENOMEM);
162
163 s->imag = av_calloc(inlink->ch_layout.nb_channels, sizeof(*s->imag));
164 if (!s->imag)
165 return AVERROR(ENOMEM);
166
167 args = av_strdup(s->real_str);
168 if (!args)
169 return AVERROR(ENOMEM);
170
171 for (ch = 0; ch < inlink->ch_layout.nb_channels; ch++) {
172 char *arg = av_strtok(ch == 0 ? args : NULL, "|", &saveptr);
173
174 ret = av_expr_parse(&s->real[ch], arg ? arg : last_expr, var_names,
175 NULL, NULL, func2_names, func2, 0, ctx);
176 if (ret < 0)
177 goto fail;
178 if (arg)
179 last_expr = arg;
180 s->nb_exprs++;
181 }
182
183 av_freep(&args);
184
185 args = av_strdup(s->img_str ? s->img_str : s->real_str);
186 if (!args)
187 return AVERROR(ENOMEM);
188
189 saveptr = NULL;
190 last_expr = "1";
191 for (ch = 0; ch < inlink->ch_layout.nb_channels; ch++) {
192 char *arg = av_strtok(ch == 0 ? args : NULL, "|", &saveptr);
193
194 ret = av_expr_parse(&s->imag[ch], arg ? arg : last_expr, var_names,
195 NULL, NULL, func2_names, func2, 0, ctx);
196 if (ret < 0)
197 goto fail;
198 if (arg)
199 last_expr = arg;
200 }
201
202 av_freep(&args);
203
204 s->window_func_lut = av_realloc_f(s->window_func_lut, s->window_size,
205 sizeof(*s->window_func_lut));
206 if (!s->window_func_lut)
207 return AVERROR(ENOMEM);
208 generate_window_func(s->window_func_lut, s->window_size, s->win_func, &overlap);
209 for (int i = 0; i < s->window_size; i++)
210 s->window_func_lut[i] = sqrtf(s->window_func_lut[i] / s->window_size);
211 if (s->overlap == 1)
212 s->overlap = overlap;
213
214 s->hop_size = s->window_size * (1 - s->overlap);
215 if (s->hop_size <= 0)
216 return AVERROR(EINVAL);
217
218 s->window = ff_get_audio_buffer(inlink, s->window_size * 2);
219 if (!s->window)
220 return AVERROR(ENOMEM);
221
222 s->buffer = ff_get_audio_buffer(inlink, s->window_size * 2);
223 if (!s->buffer)
224 return AVERROR(ENOMEM);
225
226 fail:
227 av_freep(&args);
228
229 return ret;
230 }
231
tx_channel(AVFilterContext * ctx,void * arg,int jobnr,int nb_jobs)232 static int tx_channel(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
233 {
234 AFFTFiltContext *s = ctx->priv;
235 const int channels = s->channels;
236 const int start = (channels * jobnr) / nb_jobs;
237 const int end = (channels * (jobnr+1)) / nb_jobs;
238
239 for (int ch = start; ch < end; ch++) {
240 AVComplexFloat *fft_in = s->fft_in[ch];
241 AVComplexFloat *fft_out = s->fft_out[ch];
242
243 s->tx_fn(s->fft[ch], fft_out, fft_in, sizeof(float));
244 }
245
246 return 0;
247 }
248
filter_channel(AVFilterContext * ctx,void * arg,int jobnr,int nb_jobs)249 static int filter_channel(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
250 {
251 AFFTFiltContext *s = ctx->priv;
252 const int window_size = s->window_size;
253 const float *window_lut = s->window_func_lut;
254 const float f = sqrtf(1.f - s->overlap);
255 const int channels = s->channels;
256 const int start = (channels * jobnr) / nb_jobs;
257 const int end = (channels * (jobnr+1)) / nb_jobs;
258 double values[VAR_VARS_NB];
259
260 memcpy(values, arg, sizeof(values));
261
262 for (int ch = start; ch < end; ch++) {
263 AVComplexFloat *fft_out = s->fft_out[ch];
264 AVComplexFloat *fft_temp = s->fft_temp[ch];
265 float *buf = (float *)s->buffer->extended_data[ch];
266
267 values[VAR_CHANNEL] = ch;
268
269 if (ctx->is_disabled) {
270 for (int n = 0; n < window_size; n++) {
271 fft_temp[n].re = fft_out[n].re;
272 fft_temp[n].im = fft_out[n].im;
273 }
274 } else {
275 for (int n = 0; n <= window_size / 2; n++) {
276 float fr, fi;
277
278 values[VAR_BIN] = n;
279 values[VAR_REAL] = fft_out[n].re;
280 values[VAR_IMAG] = fft_out[n].im;
281
282 fr = av_expr_eval(s->real[ch], values, s);
283 fi = av_expr_eval(s->imag[ch], values, s);
284
285 fft_temp[n].re = fr;
286 fft_temp[n].im = fi;
287 }
288
289 for (int n = window_size / 2 + 1, x = window_size / 2 - 1; n < window_size; n++, x--) {
290 fft_temp[n].re = fft_temp[x].re;
291 fft_temp[n].im = -fft_temp[x].im;
292 }
293 }
294
295 s->itx_fn(s->ifft[ch], fft_out, fft_temp, sizeof(float));
296
297 memmove(buf, buf + s->hop_size, window_size * sizeof(float));
298 for (int i = 0; i < window_size; i++)
299 buf[i] += fft_out[i].re * window_lut[i] * f;
300 }
301
302 return 0;
303 }
304
filter_frame(AVFilterLink * inlink,AVFrame * in)305 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
306 {
307 AVFilterContext *ctx = inlink->dst;
308 AVFilterLink *outlink = ctx->outputs[0];
309 AFFTFiltContext *s = ctx->priv;
310 const int window_size = s->window_size;
311 const float *window_lut = s->window_func_lut;
312 double values[VAR_VARS_NB];
313 int ch, n, ret;
314 AVFrame *out;
315
316 for (ch = 0; ch < inlink->ch_layout.nb_channels; ch++) {
317 const int offset = s->window_size - s->hop_size;
318 float *src = (float *)s->window->extended_data[ch];
319 AVComplexFloat *fft_in = s->fft_in[ch];
320
321 memmove(src, &src[s->hop_size], offset * sizeof(float));
322 memcpy(&src[offset], in->extended_data[ch], in->nb_samples * sizeof(float));
323 memset(&src[offset + in->nb_samples], 0, (s->hop_size - in->nb_samples) * sizeof(float));
324
325 for (n = 0; n < window_size; n++) {
326 fft_in[n].re = src[n] * window_lut[n];
327 fft_in[n].im = 0;
328 }
329 }
330
331 values[VAR_PTS] = in->pts;
332 values[VAR_SAMPLE_RATE] = inlink->sample_rate;
333 values[VAR_NBBINS] = window_size / 2;
334 values[VAR_CHANNELS] = inlink->ch_layout.nb_channels;
335
336 ff_filter_execute(ctx, tx_channel, NULL, NULL,
337 FFMIN(s->channels, ff_filter_get_nb_threads(ctx)));
338
339 ff_filter_execute(ctx, filter_channel, values, NULL,
340 FFMIN(s->channels, ff_filter_get_nb_threads(ctx)));
341
342 out = ff_get_audio_buffer(outlink, s->hop_size);
343 if (!out) {
344 ret = AVERROR(ENOMEM);
345 goto fail;
346 }
347
348 out->pts = in->pts;
349 out->nb_samples = in->nb_samples;
350
351 for (ch = 0; ch < inlink->ch_layout.nb_channels; ch++) {
352 float *dst = (float *)out->extended_data[ch];
353 float *buf = (float *)s->buffer->extended_data[ch];
354
355 memcpy(dst, buf, s->hop_size * sizeof(float));
356 }
357
358 ret = ff_filter_frame(outlink, out);
359 if (ret < 0)
360 goto fail;
361
362 fail:
363 av_frame_free(&in);
364 return ret < 0 ? ret : 0;
365 }
366
activate(AVFilterContext * ctx)367 static int activate(AVFilterContext *ctx)
368 {
369 AVFilterLink *inlink = ctx->inputs[0];
370 AVFilterLink *outlink = ctx->outputs[0];
371 AFFTFiltContext *s = ctx->priv;
372 AVFrame *in = NULL;
373 int ret = 0, status;
374 int64_t pts;
375
376 FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink);
377
378 ret = ff_inlink_consume_samples(inlink, s->hop_size, s->hop_size, &in);
379 if (ret < 0)
380 return ret;
381
382 if (ret > 0)
383 ret = filter_frame(inlink, in);
384 if (ret < 0)
385 return ret;
386
387 if (ff_inlink_acknowledge_status(inlink, &status, &pts)) {
388 ff_outlink_set_status(outlink, status, pts);
389 return 0;
390 }
391
392 FF_FILTER_FORWARD_WANTED(outlink, inlink);
393
394 return FFERROR_NOT_READY;
395 }
396
uninit(AVFilterContext * ctx)397 static av_cold void uninit(AVFilterContext *ctx)
398 {
399 AFFTFiltContext *s = ctx->priv;
400 int i;
401
402
403 for (i = 0; i < s->channels; i++) {
404 if (s->ifft)
405 av_tx_uninit(&s->ifft[i]);
406 if (s->fft)
407 av_tx_uninit(&s->fft[i]);
408 if (s->fft_in)
409 av_freep(&s->fft_in[i]);
410 if (s->fft_out)
411 av_freep(&s->fft_out[i]);
412 if (s->fft_temp)
413 av_freep(&s->fft_temp[i]);
414 }
415
416 av_freep(&s->fft);
417 av_freep(&s->ifft);
418 av_freep(&s->fft_in);
419 av_freep(&s->fft_out);
420 av_freep(&s->fft_temp);
421
422 for (i = 0; i < s->nb_exprs; i++) {
423 av_expr_free(s->real[i]);
424 av_expr_free(s->imag[i]);
425 }
426
427 av_freep(&s->real);
428 av_freep(&s->imag);
429 av_frame_free(&s->buffer);
430 av_frame_free(&s->window);
431 av_freep(&s->window_func_lut);
432 }
433
434 static const AVFilterPad inputs[] = {
435 {
436 .name = "default",
437 .type = AVMEDIA_TYPE_AUDIO,
438 .config_props = config_input,
439 },
440 };
441
442 static const AVFilterPad outputs[] = {
443 {
444 .name = "default",
445 .type = AVMEDIA_TYPE_AUDIO,
446 },
447 };
448
449 const AVFilter ff_af_afftfilt = {
450 .name = "afftfilt",
451 .description = NULL_IF_CONFIG_SMALL("Apply arbitrary expressions to samples in frequency domain."),
452 .priv_size = sizeof(AFFTFiltContext),
453 .priv_class = &afftfilt_class,
454 FILTER_INPUTS(inputs),
455 FILTER_OUTPUTS(outputs),
456 FILTER_SINGLE_SAMPLEFMT(AV_SAMPLE_FMT_FLTP),
457 .activate = activate,
458 .uninit = uninit,
459 .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL |
460 AVFILTER_FLAG_SLICE_THREADS,
461 };
462