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 static unsigned int get_long_le(unsigned char *p);
  31 static bool get_wave_metadata(FILE *fd, int *numchan, int *bps, int *sr, int *numsamples);
  32
  33 /* Read an unaligned 32-bit little endian long from buffer. */
  34 unsigned int get_long_le(unsigned char *p)
  35 {
  36     return p[0] | (p[1] << 8) | (p[2] << 16) | (p[3] << 24);
  37 }
  38
  39 void put_ushort_le(unsigned short x, unsigned char *out)
  40 {
  41     out[0] = x & 0xff;
  42     out[1] = x >> 8;
  43 }
  44
  45 void put_uint_le(unsigned int x, unsigned char *out)
  46 {
  47     out[0] = x & 0xff;
  48     out[1] = (x >> 8) & 0xff;
  49     out[2] = (x >> 16) & 0xff;
  50     out[3] = x >> 24;
  51 }
  52
  53
  54
  55 bool get_wave_metadata(FILE *fd, int *numchan, int *bps, int *sr, int *numsamples)
  56 {
  57     unsigned char buf[1024];
  58     unsigned long totalsamples = 0;
  59     unsigned long channels = 0;
  60     unsigned long bitspersample = 0;
  61     unsigned long numbytes = 0;
  62     size_t read_bytes;
  63     int i;
  64
  65     if ((read_bytes = fread(buf, 1, 12, fd)) < 12)
  66         return false;
  67
  68     if ((memcmp(buf, "RIFF",4) != 0) || (memcmp(&buf[8], "WAVE", 4) != 0))
  69         return false;
  70
  71     /* iterate over WAVE chunks until 'data' chunk */
  72     while (1) {
  73         /* get chunk header */
  74         if ((read_bytes = fread(buf, 1, 8, fd)) < 8)
  75             return false;
  76
  77         /* chunkSize */
  78         i = get_long_le(&buf[4]);
  79
  80         if (memcmp(buf, "fmt ", 4) == 0) {
  81             /* get rest of chunk */
  82             if ((read_bytes = fread(buf, 1, 16, fd)) < 16)
  83                 return false;
  84
  85             i -= 16;
  86
  87             channels = *numchan = buf[2] | (buf[3] << 8);
  88             *sr = get_long_le(&buf[4]);
  89             /* wBitsPerSample */
  90             bitspersample = *bps = buf[14] | (buf[15] << 8);
  91         } else if (memcmp(buf, "data", 4) == 0) {
  92             numbytes = i;
  93             break;
  94         } else if (memcmp(buf, "fact", 4) == 0) {
  95             /* dwSampleLength */
  96             if (i >= 4) {
  97                 /* get rest of chunk */
  98                 if ((read_bytes = fread(buf, 1, 4, fd)) < 4)
  99                     return false;
 100
 101                 i -= 4;
 102                 totalsamples = get_long_le(buf);
 103             }
 104         }
 105
 106         /* seek to next chunk (even chunk sizes must be padded) */
 107         if (i & 0x01)
 108             i++;
 109
 110         if (fseek(fd, i, SEEK_CUR) < 0)
 111             return false;
 112     }
 113
 114     if ((numbytes == 0) || (channels == 0))
 115         return false;
 116
 117     if (totalsamples == 0) {
 118         /* for PCM only */
 119         totalsamples = numbytes/((((bitspersample - 1) / 8) + 1)*channels);
 120     }
 121     *numsamples = totalsamples;
 122     return true;
 123 }
 124
 125 /* We'll eat an entire WAV file here, and encode it with Speex, packing the
 126  * bits as tightly as we can. Output is completely raw, with absolutely
 127  * nothing to identify the contents. Files are left open, so remember to close
 128  * them.
 129  */
 130 bool encode_file(FILE *fin, FILE *fout, float quality, int complexity,
 131                  bool narrowband, float volume, char *errstr, size_t errlen)
 132 {
 133     spx_int16_t *in = NULL, *inpos;
 134     spx_int16_t enc_buf[640]; /* Max frame size */
 135     char cbits[200];
 136     void *st = NULL;
 137     SpeexResamplerState *resampler = NULL;
 138     SpeexBits bits;
 139     int i, tmp, target_sr, numchan, bps, sr, numsamples, frame_size, lookahead;
 140     int nbytes;
 141     bool ret = true;
 142     int a;
 143
 144     if (!get_wave_metadata(fin, &numchan, &bps, &sr, &numsamples)) {
 145         snprintf(errstr, errlen, "invalid WAV file");
 146         return false;
 147     }
 148     if (numchan != 1) {
 149         snprintf(errstr, errlen, "input file must be mono");
 150         return false;
 151     }
 152     if (bps != 16) {
 153         snprintf(errstr, errlen, "samples must be 16 bit");
 154         return false;
 155     }
 156
 157     /* Allocate an encoder of specified type, defaults to wideband */
 158     st = speex_encoder_init(narrowband ? &speex_nb_mode : &speex_wb_mode);
 159     if (narrowband)
 160         target_sr = 8000;
 161     else
 162         target_sr = 16000;
 163     speex_bits_init(&bits);
 164
 165     /* VBR */
 166     tmp = 1;
 167     speex_encoder_ctl(st, SPEEX_SET_VBR, &tmp);
 168     /* Quality, 0-10 */
 169     speex_encoder_ctl(st, SPEEX_SET_VBR_QUALITY, &quality);
 170     /* Complexity, 0-10 */
 171     speex_encoder_ctl(st, SPEEX_SET_COMPLEXITY, &complexity);
 172     speex_encoder_ctl(st, SPEEX_GET_FRAME_SIZE, &frame_size);
 173     speex_encoder_ctl(st, SPEEX_GET_LOOKAHEAD, &lookahead);
 174
 175     /* Read input samples into a buffer */
 176     in = calloc(numsamples + lookahead, sizeof(spx_int16_t));
 177     if (in == NULL) {
 178         snprintf(errstr, errlen, "could not allocate clip memory");
 179         ret = false;
 180         goto finish;
 181     }
 182     if (fread(in, 2, numsamples, fin) != numsamples) {
 183         snprintf(errstr, errlen, "could not read input file data");
 184         ret = false;
 185         goto finish;
 186     }
 187 #if defined(__BIG_ENDIAN__)
 188     /* byteswap read bytes to host endianess. */
 189     a = numsamples;
 190     while(a--) {
 191         *(in + a) = ((unsigned short)(*(in + a)) >> 8) & 0x00ff
 192                   | ((unsigned short)(*(in + a)) << 8) & 0xff00;
 193     }
 194 #endif
 195
 196     if (volume != 1.0f) {
 197         for (i = 0; i < numsamples; ++i)
 198             in[i] *= volume;
 199     }
 200
 201     if (sr != target_sr) {
 202         resampler = speex_resampler_init(1, sr, target_sr, 10, NULL);
 203         speex_resampler_skip_zeros(resampler);
 204     }
 205
 206     /* There will be 'lookahead' samples of zero at the end of the array, to
 207      * make sure the Speex encoder is allowed to spit out all its data at clip
 208      * end */
 209     numsamples += lookahead;
 210
 211     inpos = in;
 212     while (numsamples > 0) {
 213         int samples = frame_size;
 214
 215         /* Check if we need to resample */
 216         if (sr != target_sr) {
 217             spx_uint32_t in_len = numsamples, out_len = frame_size;
 218             double resample_factor = (double)sr/(double)target_sr;
 219             /* Calculate how many input samples are needed for one full frame
 220              * out, and add some, just in case. */
 221             spx_uint32_t samples_in = frame_size*resample_factor + 50;
 222
 223             /* Limit this or resampler will try to allocate it all on stack */
 224             if (in_len > samples_in)
 225                 in_len = samples_in;
 226             speex_resampler_process_int(resampler, 0, inpos, &in_len,
 227                                         enc_buf, &out_len);
 228             inpos += in_len;
 229             samples = out_len;
 230             numsamples -= in_len;
 231         } else {
 232             if (samples > numsamples)
 233                 samples = numsamples;
 234             memcpy(enc_buf, inpos, samples*2);
 235             inpos += frame_size;
 236             numsamples -= frame_size;
 237         }
 238         /* Pad out with zeros if we didn't fill all input */
 239         memset(enc_buf + samples, 0, (frame_size - samples)*2);
 240
 241         if (speex_encode_int(st, enc_buf, &bits) < 0) {
 242             snprintf(errstr, errlen, "encoder error");
 243             ret = false;
 244             goto finish;
 245         }
 246
 247         /* Copy the bits to an array of char that can be written */
 248         nbytes = speex_bits_write_whole_bytes(&bits, cbits, 200);
 249
 250         /* Write the compressed data */
 251         if (fwrite(cbits, 1, nbytes, fout) != nbytes) {
 252             snprintf(errstr, errlen, "could not write output data");
 253             ret = false;
 254             goto finish;
 255         }
 256     }
 257     /* Squeeze out the last bits */
 258     nbytes = speex_bits_write(&bits, cbits, 200);
 259     if (fwrite(cbits, 1, nbytes, fout) != nbytes) {
 260         snprintf(errstr, errlen, "could not write output data");
 261         ret = false;
 262     }
 263
 264 finish:
 265     if (st != NULL)
 266         speex_encoder_destroy(st);
 267     speex_bits_destroy(&bits);
 268     if (resampler != NULL)
 269         speex_resampler_destroy(resampler);
 270     if (in != NULL)
 271         free(in);
 272     return ret;
 273 }
 274
 275