0
|
1 /*
|
|
2 * Sample rate convertion for both audio and video
|
|
3 * Copyright (c) 2000 Gerard Lantau.
|
|
4 *
|
|
5 * This program is free software; you can redistribute it and/or modify
|
|
6 * it under the terms of the GNU General Public License as published by
|
|
7 * the Free Software Foundation; either version 2 of the License, or
|
|
8 * (at your option) any later version.
|
|
9 *
|
|
10 * This program is distributed in the hope that it will be useful,
|
|
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
13 * GNU General Public License for more details.
|
|
14 *
|
|
15 * You should have received a copy of the GNU General Public License
|
|
16 * along with this program; if not, write to the Free Software
|
|
17 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
|
|
18 */
|
|
19 #include <stdlib.h>
|
|
20 #include <stdio.h>
|
|
21 #include <string.h>
|
|
22 #include <math.h>
|
|
23 #include "avcodec.h"
|
|
24
|
|
25 #define NDEBUG
|
|
26 #include <assert.h>
|
|
27
|
|
28 typedef struct {
|
|
29 /* fractional resampling */
|
|
30 UINT32 incr; /* fractional increment */
|
|
31 UINT32 frac;
|
|
32 int last_sample;
|
|
33 /* integer down sample */
|
|
34 int iratio; /* integer divison ratio */
|
|
35 int icount, isum;
|
|
36 int inv;
|
|
37 } ReSampleChannelContext;
|
|
38
|
|
39 struct ReSampleContext {
|
|
40 ReSampleChannelContext channel_ctx[2];
|
|
41 float ratio;
|
|
42 /* channel convert */
|
|
43 int input_channels, output_channels, filter_channels;
|
|
44 };
|
|
45
|
|
46
|
|
47 #define FRAC_BITS 16
|
|
48 #define FRAC (1 << FRAC_BITS)
|
|
49
|
|
50 static void init_mono_resample(ReSampleChannelContext *s, float ratio)
|
|
51 {
|
|
52 ratio = 1.0 / ratio;
|
|
53 s->iratio = (int)floor(ratio);
|
|
54 if (s->iratio == 0)
|
|
55 s->iratio = 1;
|
|
56 s->incr = (int)((ratio / s->iratio) * FRAC);
|
|
57 s->frac = 0;
|
|
58 s->last_sample = 0;
|
|
59 s->icount = s->iratio;
|
|
60 s->isum = 0;
|
|
61 s->inv = (FRAC / s->iratio);
|
|
62 }
|
|
63
|
|
64 /* fractional audio resampling */
|
|
65 static int fractional_resample(ReSampleChannelContext *s, short *output, short *input, int nb_samples)
|
|
66 {
|
|
67 unsigned int frac, incr;
|
|
68 int l0, l1;
|
|
69 short *q, *p, *pend;
|
|
70
|
|
71 l0 = s->last_sample;
|
|
72 incr = s->incr;
|
|
73 frac = s->frac;
|
|
74
|
|
75 p = input;
|
|
76 pend = input + nb_samples;
|
|
77 q = output;
|
|
78
|
|
79 l1 = *p++;
|
|
80 for(;;) {
|
|
81 /* interpolate */
|
|
82 *q++ = (l0 * (FRAC - frac) + l1 * frac) >> FRAC_BITS;
|
|
83 frac = frac + s->incr;
|
|
84 while (frac >= FRAC) {
|
|
85 if (p >= pend)
|
|
86 goto the_end;
|
|
87 frac -= FRAC;
|
|
88 l0 = l1;
|
|
89 l1 = *p++;
|
|
90 }
|
|
91 }
|
|
92 the_end:
|
|
93 s->last_sample = l1;
|
|
94 s->frac = frac;
|
|
95 return q - output;
|
|
96 }
|
|
97
|
|
98 static int integer_downsample(ReSampleChannelContext *s, short *output, short *input, int nb_samples)
|
|
99 {
|
|
100 short *q, *p, *pend;
|
|
101 int c, sum;
|
|
102
|
|
103 p = input;
|
|
104 pend = input + nb_samples;
|
|
105 q = output;
|
|
106
|
|
107 c = s->icount;
|
|
108 sum = s->isum;
|
|
109
|
|
110 for(;;) {
|
|
111 sum += *p++;
|
|
112 if (--c == 0) {
|
|
113 *q++ = (sum * s->inv) >> FRAC_BITS;
|
|
114 c = s->iratio;
|
|
115 sum = 0;
|
|
116 }
|
|
117 if (p >= pend)
|
|
118 break;
|
|
119 }
|
|
120 s->isum = sum;
|
|
121 s->icount = c;
|
|
122 return q - output;
|
|
123 }
|
|
124
|
|
125 /* n1: number of samples */
|
|
126 static void stereo_to_mono(short *output, short *input, int n1)
|
|
127 {
|
|
128 short *p, *q;
|
|
129 int n = n1;
|
|
130
|
|
131 p = input;
|
|
132 q = output;
|
|
133 while (n >= 4) {
|
|
134 q[0] = (p[0] + p[1]) >> 1;
|
|
135 q[1] = (p[2] + p[3]) >> 1;
|
|
136 q[2] = (p[4] + p[5]) >> 1;
|
|
137 q[3] = (p[6] + p[7]) >> 1;
|
|
138 q += 4;
|
|
139 p += 8;
|
|
140 n -= 4;
|
|
141 }
|
|
142 while (n > 0) {
|
|
143 q[0] = (p[0] + p[1]) >> 1;
|
|
144 q++;
|
|
145 p += 2;
|
|
146 n--;
|
|
147 }
|
|
148 }
|
|
149
|
|
150 /* n1: number of samples */
|
|
151 static void mono_to_stereo(short *output, short *input, int n1)
|
|
152 {
|
|
153 short *p, *q;
|
|
154 int n = n1;
|
|
155 int v;
|
|
156
|
|
157 p = input;
|
|
158 q = output;
|
|
159 while (n >= 4) {
|
|
160 v = p[0]; q[0] = v; q[1] = v;
|
|
161 v = p[1]; q[2] = v; q[3] = v;
|
|
162 v = p[2]; q[4] = v; q[5] = v;
|
|
163 v = p[3]; q[6] = v; q[7] = v;
|
|
164 q += 8;
|
|
165 p += 4;
|
|
166 n -= 4;
|
|
167 }
|
|
168 while (n > 0) {
|
|
169 v = p[0]; q[0] = v; q[1] = v;
|
|
170 q += 2;
|
|
171 p += 1;
|
|
172 n--;
|
|
173 }
|
|
174 }
|
|
175
|
|
176 /* XXX: should use more abstract 'N' channels system */
|
|
177 static void stereo_split(short *output1, short *output2, short *input, int n)
|
|
178 {
|
|
179 int i;
|
|
180
|
|
181 for(i=0;i<n;i++) {
|
|
182 *output1++ = *input++;
|
|
183 *output2++ = *input++;
|
|
184 }
|
|
185 }
|
|
186
|
|
187 static void stereo_mux(short *output, short *input1, short *input2, int n)
|
|
188 {
|
|
189 int i;
|
|
190
|
|
191 for(i=0;i<n;i++) {
|
|
192 *output++ = *input1++;
|
|
193 *output++ = *input2++;
|
|
194 }
|
|
195 }
|
|
196
|
|
197 static int mono_resample(ReSampleChannelContext *s, short *output, short *input, int nb_samples)
|
|
198 {
|
|
199 short buf1[nb_samples];
|
|
200 short *buftmp;
|
|
201
|
|
202 /* first downsample by an integer factor with averaging filter */
|
|
203 if (s->iratio > 1) {
|
|
204 buftmp = buf1;
|
|
205 nb_samples = integer_downsample(s, buftmp, input, nb_samples);
|
|
206 } else {
|
|
207 buftmp = input;
|
|
208 }
|
|
209
|
|
210 /* then do a fractional resampling with linear interpolation */
|
|
211 if (s->incr != FRAC) {
|
|
212 nb_samples = fractional_resample(s, output, buftmp, nb_samples);
|
|
213 } else {
|
|
214 memcpy(output, buftmp, nb_samples * sizeof(short));
|
|
215 }
|
|
216 return nb_samples;
|
|
217 }
|
|
218
|
|
219 ReSampleContext *audio_resample_init(int output_channels, int input_channels,
|
|
220 int output_rate, int input_rate)
|
|
221 {
|
|
222 ReSampleContext *s;
|
|
223 int i;
|
|
224
|
|
225 if (output_channels > 2 || input_channels > 2)
|
|
226 return NULL;
|
|
227
|
|
228 s = av_mallocz(sizeof(ReSampleContext));
|
|
229 if (!s)
|
|
230 return NULL;
|
|
231
|
|
232 s->ratio = (float)output_rate / (float)input_rate;
|
|
233
|
|
234 s->input_channels = input_channels;
|
|
235 s->output_channels = output_channels;
|
|
236
|
|
237 s->filter_channels = s->input_channels;
|
|
238 if (s->output_channels < s->filter_channels)
|
|
239 s->filter_channels = s->output_channels;
|
|
240
|
|
241 for(i=0;i<s->filter_channels;i++) {
|
|
242 init_mono_resample(&s->channel_ctx[i], s->ratio);
|
|
243 }
|
|
244 return s;
|
|
245 }
|
|
246
|
|
247 /* resample audio. 'nb_samples' is the number of input samples */
|
|
248 /* XXX: optimize it ! */
|
|
249 /* XXX: do it with polyphase filters, since the quality here is
|
|
250 HORRIBLE. Return the number of samples available in output */
|
|
251 int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples)
|
|
252 {
|
|
253 int i, nb_samples1;
|
|
254 short bufin[2][nb_samples];
|
|
255 short bufout[2][(int)(nb_samples * s->ratio) + 16]; /* make some zoom to avoid round pb */
|
|
256 short *buftmp2[2], *buftmp3[2];
|
|
257
|
|
258 if (s->input_channels == s->output_channels && s->ratio == 1.0) {
|
|
259 /* nothing to do */
|
|
260 memcpy(output, input, nb_samples * s->input_channels * sizeof(short));
|
|
261 return nb_samples;
|
|
262 }
|
|
263
|
|
264 if (s->input_channels == 2 &&
|
|
265 s->output_channels == 1) {
|
|
266 buftmp2[0] = bufin[0];
|
|
267 buftmp3[0] = output;
|
|
268 stereo_to_mono(buftmp2[0], input, nb_samples);
|
|
269 } else if (s->output_channels == 2 && s->input_channels == 1) {
|
|
270 buftmp2[0] = input;
|
|
271 buftmp3[0] = bufout[0];
|
|
272 } else if (s->output_channels == 2) {
|
|
273 buftmp2[0] = bufin[0];
|
|
274 buftmp2[1] = bufin[1];
|
|
275 buftmp3[0] = bufout[0];
|
|
276 buftmp3[1] = bufout[1];
|
|
277 stereo_split(buftmp2[0], buftmp2[1], input, nb_samples);
|
|
278 } else {
|
|
279 buftmp2[0] = input;
|
|
280 buftmp3[0] = output;
|
|
281 }
|
|
282
|
|
283 /* resample each channel */
|
|
284 nb_samples1 = 0; /* avoid warning */
|
|
285 for(i=0;i<s->filter_channels;i++) {
|
|
286 nb_samples1 = mono_resample(&s->channel_ctx[i], buftmp3[i], buftmp2[i], nb_samples);
|
|
287 }
|
|
288
|
|
289 if (s->output_channels == 2 && s->input_channels == 1) {
|
|
290 mono_to_stereo(output, buftmp3[0], nb_samples1);
|
|
291 } else if (s->output_channels == 2) {
|
|
292 stereo_mux(output, buftmp3[0], buftmp3[1], nb_samples1);
|
|
293 }
|
|
294
|
|
295 return nb_samples1;
|
|
296 }
|
|
297
|
|
298 void audio_resample_close(ReSampleContext *s)
|
|
299 {
|
|
300 free(s);
|
|
301 }
|