tools/rbspeex/rbspeex.c

   1 /**************************************************************************
   2  *             __________               __   ___.
   3  *   Open      \______   \ ____   ____ |  | _\_ |__   _______  ___
   4  *   Source     |       _//  _ \_/ ___\|  |/ /| __ \ /  _ \  \/  /
   5  *   Jukebox    |    |   (  <_> )  \___|    < | \_\ (  <_> > <  <
   6  *   Firmware   |____|_  /\____/ \___  >__|_ \|___  /\____/__/\_ \
   7  *                     \/            \/     \/    \/            \/
   8  *
   9  * Copyright (C) 2007 Thom Johansen
  10  *
  11  * This program is free software; you can redistribute it and/or
  12  * modify it under the terms of the GNU General Public License
  13  * as published by the Free Software Foundation; either version 2
  14  * of the License, or (at your option) any later version.
  15  *
  16  * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
  17  * KIND, either express or implied.
  18  *
  19  ***************************************************************************/
  20
  21 #include <speex/speex.h>
  22 #include <speex/speex_resampler.h>
  23 #include <stdio.h>
  24 #include <stdlib.h>
  25 #include <string.h>
  26 #include <stdbool.h>
  27
  28 #include "rbspeex.h"
  29
  30 /* Read an unaligned 32-bit little endian long from buffer. */
  31 unsigned int get_long_le(unsigned char *p)
  32 {
  33     return p[0] | (p[1] << 8) | (p[2] << 16) | (p[3] << 24);
  34 }
  35
  36 void put_ushort_le(unsigned short x, unsigned char *out)
  37 {
  38     out[0] = x & 0xff;
  39     out[1] = x >> 8;
  40 }
  41
  42 void put_uint_le(unsigned int x, unsigned char *out)
  43 {
  44     out[0] = x & 0xff;
  45     out[1] = (x >> 8) & 0xff;
  46     out[2] = (x >> 16) & 0xff;
  47     out[3] = x >> 24;
  48 }
  49
  50
  51
  52 bool get_wave_metadata(FILE *fd, int *numchan, int *bps, int *sr, int *numsamples)
  53 {
  54     unsigned char buf[1024];
  55     unsigned long totalsamples = 0;
  56     unsigned long channels = 0;
  57     unsigned long bitspersample = 0;
  58     unsigned long numbytes = 0;
  59     size_t read_bytes;
  60     int i;
  61
  62     if ((read_bytes = fread(buf, 1, 12, fd)) < 12)
  63         return false;
  64
  65     if ((memcmp(buf, "RIFF",4) != 0) || (memcmp(&buf[8], "WAVE", 4) != 0))
  66         return false;
  67
  68     /* iterate over WAVE chunks until 'data' chunk */
  69     while (1) {
  70         /* get chunk header */
  71         if ((read_bytes = fread(buf, 1, 8, fd)) < 8)
  72             return false;
  73
  74         /* chunkSize */
  75         i = get_long_le(&buf[4]);
  76
  77         if (memcmp(buf, "fmt ", 4) == 0) {
  78             /* get rest of chunk */
  79             if ((read_bytes = fread(buf, 1, 16, fd)) < 16)
  80                 return false;
  81
  82             i -= 16;
  83
  84             channels = *numchan = buf[2] | (buf[3] << 8);
  85             *sr = get_long_le(&buf[4]);
  86             /* wBitsPerSample */
  87             bitspersample = *bps = buf[14] | (buf[15] << 8);
  88         } else if (memcmp(buf, "data", 4) == 0) {
  89             numbytes = i;
  90             break;
  91         } else if (memcmp(buf, "fact", 4) == 0) {
  92             /* dwSampleLength */
  93             if (i >= 4) {
  94                 /* get rest of chunk */
  95                 if ((read_bytes = fread(buf, 1, 4, fd)) < 4)
  96                     return false;
  97
  98                 i -= 4;
  99                 totalsamples = get_long_le(buf);
 100             }
 101         }
 102
 103         /* seek to next chunk (even chunk sizes must be padded) */
 104         if (i & 0x01)
 105             i++;
 106
 107         if (fseek(fd, i, SEEK_CUR) < 0)
 108             return false;
 109     }
 110
 111     if ((numbytes == 0) || (channels == 0))
 112         return false;
 113
 114     if (totalsamples == 0) {
 115         /* for PCM only */
 116         totalsamples = numbytes/((((bitspersample - 1) / 8) + 1)*channels);
 117     }
 118     *numsamples = totalsamples;
 119     return true;
 120 }
 121
 122 /* We'll eat an entire WAV file here, and encode it with Speex, packing the
 123  * bits as tightly as we can. Output is completely raw, with absolutely
 124  * nothing to identify the contents. Files are left open, so remember to close
 125  * them.
 126  */
 127 bool encode_file(FILE *fin, FILE *fout, float quality, int complexity,
 128                  bool narrowband, float volume, char *errstr, size_t errlen)
 129 {
 130     spx_int16_t *in = NULL, *inpos;
 131     spx_int16_t enc_buf[640]; /* Max frame size */
 132     char cbits[200];
 133     void *st = NULL;
 134     SpeexResamplerState *resampler = NULL;
 135     SpeexBits bits;
 136     int i, tmp, target_sr, numchan, bps, sr, numsamples, frame_size, lookahead;
 137     int nbytes;
 138     bool ret = true;
 139     int a;
 140
 141     if (!get_wave_metadata(fin, &numchan, &bps, &sr, &numsamples)) {
 142         snprintf(errstr, errlen, "invalid WAV file");
 143         return false;
 144     }
 145     if (numchan != 1) {
 146         snprintf(errstr, errlen, "input file must be mono");
 147         return false;
 148     }
 149     if (bps != 16) {
 150         snprintf(errstr, errlen, "samples must be 16 bit");
 151         return false;
 152     }
 153
 154     /* Allocate an encoder of specified type, defaults to wideband */
 155     st = speex_encoder_init(narrowband ? &speex_nb_mode : &speex_wb_mode);
 156     if (narrowband)
 157         target_sr = 8000;
 158     else
 159         target_sr = 16000;
 160     speex_bits_init(&bits);
 161
 162     /* VBR */
 163     tmp = 1;
 164     speex_encoder_ctl(st, SPEEX_SET_VBR, &tmp);
 165     /* Quality, 0-10 */
 166     speex_encoder_ctl(st, SPEEX_SET_VBR_QUALITY, &quality);
 167     /* Complexity, 0-10 */
 168     speex_encoder_ctl(st, SPEEX_SET_COMPLEXITY, &complexity);
 169     speex_encoder_ctl(st, SPEEX_GET_FRAME_SIZE, &frame_size);
 170     speex_encoder_ctl(st, SPEEX_GET_LOOKAHEAD, &lookahead);
 171
 172     /* Read input samples into a buffer */
 173     in = calloc(numsamples + lookahead, sizeof(spx_int16_t));
 174     if (in == NULL) {
 175         snprintf(errstr, errlen, "could not allocate clip memory");
 176         ret = false;
 177         goto finish;
 178     }
 179     if (fread(in, 2, numsamples, fin) != numsamples) {
 180         snprintf(errstr, errlen, "could not read input file data");
 181         ret = false;
 182         goto finish;
 183     }
 184 #if defined(__BIG_ENDIAN__)
 185     /* byteswap read bytes to host endianess. */
 186     a = numsamples;
 187     while(a--) {
 188         *(in + a) = ((unsigned short)(*(in + a)) >> 8) & 0x00ff
 189                   | ((unsigned short)(*(in + a)) << 8) & 0xff00;
 190     }
 191 #endif
 192
 193     if (volume != 1.0f) {
 194         for (i = 0; i < numsamples; ++i)
 195             in[i] *= volume;
 196     }
 197
 198     if (sr != target_sr) {
 199         resampler = speex_resampler_init(1, sr, target_sr, 10, NULL);
 200         speex_resampler_skip_zeros(resampler);
 201     }
 202
 203     /* There will be 'lookahead' samples of zero at the end of the array, to
 204      * make sure the Speex encoder is allowed to spit out all its data at clip
 205      * end */
 206     numsamples += lookahead;
 207
 208     inpos = in;
 209     while (numsamples > 0) {
 210         int samples = frame_size;
 211
 212         /* Check if we need to resample */
 213         if (sr != target_sr) {
 214             spx_uint32_t in_len = numsamples, out_len = frame_size;
 215             double resample_factor = (double)sr/(double)target_sr;
 216             /* Calculate how many input samples are needed for one full frame
 217              * out, and add some, just in case. */
 218             spx_uint32_t samples_in = frame_size*resample_factor + 50;
 219
 220             /* Limit this or resampler will try to allocate it all on stack */
 221             if (in_len > samples_in)
 222                 in_len = samples_in;
 223             speex_resampler_process_int(resampler, 0, inpos, &in_len,
 224                                         enc_buf, &out_len);
 225             inpos += in_len;
 226             samples = out_len;
 227             numsamples -= in_len;
 228         } else {
 229             if (samples > numsamples)
 230                 samples = numsamples;
 231             memcpy(enc_buf, inpos, samples*2);
 232             inpos += frame_size;
 233             numsamples -= frame_size;
 234         }
 235         /* Pad out with zeros if we didn't fill all input */
 236         memset(enc_buf + samples, 0, (frame_size - samples)*2);
 237
 238         if (speex_encode_int(st, enc_buf, &bits) < 0) {
 239             snprintf(errstr, errlen, "encoder error");
 240             ret = false;
 241             goto finish;
 242         }
 243
 244         /* Copy the bits to an array of char that can be written */
 245         nbytes = speex_bits_write_whole_bytes(&bits, cbits, 200);
 246
 247         /* Write the compressed data */
 248         if (fwrite(cbits, 1, nbytes, fout) != nbytes) {
 249             snprintf(errstr, errlen, "could not write output data");
 250             ret = false;
 251             goto finish;
 252         }
 253     }
 254     /* Squeeze out the last bits */
 255     nbytes = speex_bits_write(&bits, cbits, 200);
 256     if (fwrite(cbits, 1, nbytes, fout) != nbytes) {
 257         snprintf(errstr, errlen, "could not write output data");
 258         ret = false;
 259     }
 260
 261 finish:
 262     if (st != NULL)
 263         speex_encoder_destroy(st);
 264     speex_bits_destroy(&bits);
 265     if (resampler != NULL)
 266         speex_resampler_destroy(resampler);
 267     if (in != NULL)
 268         free(in);
 269     return ret;
 270 }
 271
 272