mame/src/tools/ldresample.cpp
Miodrag Milanovic ddb290d5f6 NOTICE (TYPE NAME CONSOLIDATION)
Use standard uint64_t, uint32_t, uint16_t or uint8_t instead of UINT64, UINT32, UINT16 or UINT8
also use standard int64_t, int32_t, int16_t or int8_t instead of INT64, INT32, INT16 or INT8
2016-10-22 13:13:17 +02:00

575 lines
18 KiB
C++

// license:BSD-3-Clause
// copyright-holders:Aaron Giles
/***************************************************************************
ldresample.c
Laserdisc audio synchronizer and resampler.
****************************************************************************/
#include <stdio.h>
#include <ctype.h>
#include <stdlib.h>
#include <math.h>
#include <new>
#include <assert.h>
#include "bitmap.h"
#include "chd.h"
#include "avhuff.h"
#include "vbiparse.h"
//**************************************************************************
// CONSTANTS
//**************************************************************************
// size of window where we scan ahead to find maximum; this should be large enough to
// catch peaks of even slow waves
const uint32_t MAXIMUM_WINDOW_SIZE = 40;
// number of standard deviations away from silence that we consider a real signal
const uint32_t SIGNAL_DEVIATIONS = 100;
// number of standard deviations away from silence that we consider the start of a signal
const uint32_t SIGNAL_START_DEVIATIONS = 5;
// number of consecutive entries of signal before we consider that we found it
const uint32_t MINIMUM_SIGNAL_COUNT = 20;
//**************************************************************************
// TYPE DEFINITIONS
//**************************************************************************
struct movie_info
{
double framerate;
int iframerate;
int numfields;
int width;
int height;
int samplerate;
int channels;
int interlaced;
bitmap_yuy16 bitmap;
std::vector<int16_t> lsound;
std::vector<int16_t> rsound;
uint32_t samples;
};
// ======================> chd_resample_compressor
class chd_resample_compressor : public chd_file_compressor
{
public:
// construction/destruction
chd_resample_compressor(chd_file &source, movie_info &info, int64_t ioffset, int64_t islope)
: m_source(source),
m_info(info),
m_ioffset(ioffset),
m_islope(islope) { }
// read interface
virtual uint32_t read_data(void *_dest, uint64_t offset, uint32_t length)
{
assert(offset % m_source.hunk_bytes() == 0);
assert(length % m_source.hunk_bytes() == 0);
uint32_t startfield = offset / m_source.hunk_bytes();
uint32_t endfield = startfield + length / m_source.hunk_bytes();
uint8_t *dest = reinterpret_cast<uint8_t *>(_dest);
for (uint32_t fieldnum = startfield; fieldnum < endfield; fieldnum++)
{
generate_one_frame(dest, m_source.hunk_bytes(), fieldnum);
dest += m_source.hunk_bytes();
}
return length;
}
private:
// internal helpers
void generate_one_frame(uint8_t *dest, uint32_t datasize, uint32_t fieldnum);
// internal state
chd_file & m_source;
movie_info & m_info;
int64_t m_ioffset;
int64_t m_islope;
};
//**************************************************************************
// INLINE FUNCTIONS
//**************************************************************************
//-------------------------------------------------
// field_to_sample_number - given a field number
// compute the absolute sample number for the
// first sample of that field
//-------------------------------------------------
inline uint32_t field_to_sample_number(const movie_info &info, uint32_t field)
{
return (uint64_t(info.samplerate) * uint64_t(field) * uint64_t(1000000) + info.iframerate - 1) / uint64_t(info.iframerate);
}
//-------------------------------------------------
// sample_number_to_field - given a sample number
// compute the field where it is located and
// the offset within the field
//-------------------------------------------------
inline uint32_t sample_number_to_field(const movie_info &info, uint32_t samplenum, uint32_t &offset)
{
uint32_t guess = (uint64_t(samplenum) * uint64_t(info.iframerate) + (uint64_t(info.samplerate) * uint64_t(1000000) - 1)) / (uint64_t(info.samplerate) * uint64_t(1000000));
while (1)
{
uint32_t fieldstart = field_to_sample_number(info, guess);
uint32_t fieldend = field_to_sample_number(info, guess + 1);
if (samplenum >= fieldstart && samplenum < fieldend)
{
offset = samplenum - fieldstart;
return guess;
}
else if (samplenum < fieldstart)
guess--;
else
guess++;
}
}
//**************************************************************************
// CHD HANDLING
//**************************************************************************
//-------------------------------------------------
// open_chd - open a CHD file and return
// information about it
//-------------------------------------------------
static chd_error open_chd(chd_file &file, const char *filename, movie_info &info)
{
// open the file
chd_error chderr = file.open(filename);
if (chderr != CHDERR_NONE)
{
fprintf(stderr, "Error opening CHD file: %s\n", chd_file::error_string(chderr));
return chderr;
}
// get the metadata
std::string metadata;
chderr = file.read_metadata(AV_METADATA_TAG, 0, metadata);
if (chderr != CHDERR_NONE)
{
fprintf(stderr, "Error getting A/V metadata: %s\n", chd_file::error_string(chderr));
return chderr;
}
// extract the info
int fps, fpsfrac, width, height, interlaced, channels, rate;
if (sscanf(metadata.c_str(), AV_METADATA_FORMAT, &fps, &fpsfrac, &width, &height, &interlaced, &channels, &rate) != 7)
{
fprintf(stderr, "Improperly formatted metadata\n");
return CHDERR_INVALID_DATA;
}
// extract movie info
info.iframerate = fps * 1000000 + fpsfrac;
info.framerate = info.iframerate / 1000000.0;
info.numfields = file.hunk_count();
info.width = width;
info.height = height;
info.interlaced = interlaced;
info.samplerate = rate;
info.channels = channels;
// allocate buffers
info.bitmap.resize(info.width, info.height);
info.lsound.resize(info.samplerate);
info.rsound.resize(info.samplerate);
return CHDERR_NONE;
}
//-------------------------------------------------
// create_chd - create a new CHD file
//-------------------------------------------------
static chd_error create_chd(chd_file_compressor &file, const char *filename, chd_file &source, const movie_info &info)
{
// create the file
chd_codec_type compression[4] = { CHD_CODEC_AVHUFF };
chd_error chderr = file.create(filename, source.logical_bytes(), source.hunk_bytes(), source.unit_bytes(), compression);
if (chderr != CHDERR_NONE)
{
fprintf(stderr, "Error creating new CHD file: %s\n", chd_file::error_string(chderr));
return chderr;
}
// clone the metadata
chderr = file.clone_all_metadata(source);
if (chderr != CHDERR_NONE)
{
fprintf(stderr, "Error cloning metadata: %s\n", chd_file::error_string(chderr));
return chderr;
}
// begin compressing
file.compress_begin();
return CHDERR_NONE;
}
//-------------------------------------------------
// read_chd - read a field from a CHD file
//-------------------------------------------------
static bool read_chd(chd_file &file, uint32_t field, movie_info &info, uint32_t soundoffs)
{
// configure the codec
avhuff_decompress_config avconfig;
avconfig.video.wrap(info.bitmap, info.bitmap.cliprect());
avconfig.maxsamples = info.lsound.size();
avconfig.actsamples = &info.samples;
avconfig.audio[0] = &info.lsound[soundoffs];
avconfig.audio[1] = &info.rsound[soundoffs];
// configure the decompressor for this field
file.codec_configure(CHD_CODEC_AVHUFF, AVHUFF_CODEC_DECOMPRESS_CONFIG, &avconfig);
// read the field
chd_error chderr = file.read_hunk(field, nullptr);
return (chderr == CHDERR_NONE);
}
//**************************************************************************
// CORE IMPLEMENTATION
//**************************************************************************
//-------------------------------------------------
// find_edge_near_field - given a field number,
// load +/- 1/2 second on either side and find
// an audio edge
//-------------------------------------------------
static bool find_edge_near_field(chd_file &srcfile, uint32_t fieldnum, movie_info &info, bool report_best_field, int32_t &delta)
{
// clear the sound buffers
memset(&info.lsound[0], 0, info.lsound.size() * 2);
memset(&info.rsound[0], 0, info.rsound.size() * 2);
// read 1 second around the target area
int fields_to_read = info.iframerate / 1000000;
int32_t firstfield = fieldnum - (fields_to_read / 2);
uint32_t targetsoundstart = 0;
uint32_t firstfieldend = 0;
uint32_t fieldstart[100];
uint32_t soundend = 0;
for (int32_t curfield = 0; curfield < fields_to_read; curfield++)
{
// remember the start of each field
fieldstart[curfield] = soundend;
// remember the sound offset where the initial fieldnum is
if (firstfield + curfield == fieldnum)
targetsoundstart = soundend;
// read the frame and samples
if (firstfield + curfield >= 0)
{
read_chd(srcfile, firstfield + curfield, info, soundend);
soundend += info.samples;
// also remember the offset at the end of the first field
if (firstfieldend == 0)
firstfieldend = soundend;
}
}
// compute absolute deltas across the samples
for (uint32_t sampnum = 0; sampnum < soundend; sampnum++)
{
info.lsound[sampnum] = labs(info.lsound[sampnum + 1] - info.lsound[sampnum]);
info.rsound[sampnum] = labs(info.rsound[sampnum + 1] - info.rsound[sampnum]);
}
// for each sample in the collection, find the highest deltas over the
// next few samples, and take the nth highest value (to remove outliers)
for (uint32_t sampnum = 0; sampnum < soundend - MAXIMUM_WINDOW_SIZE; sampnum++)
{
// scan forward over the maximum window
uint32_t lmax = 0, rmax = 0;
for (uint32_t scannum = 0; scannum < MAXIMUM_WINDOW_SIZE; scannum++)
{
if (info.lsound[sampnum + scannum] > lmax)
lmax = info.lsound[sampnum + scannum];
if (info.rsound[sampnum + scannum] > rmax)
rmax = info.rsound[sampnum + scannum];
}
// replace this sample with the maximum value found
info.lsound[sampnum] = lmax;
info.rsound[sampnum] = rmax;
}
// now compute the average over the first field, which is assumed to be silence
uint32_t firstlavg = 0;
uint32_t firstravg = 0;
for (uint32_t sampnum = 0; sampnum < firstfieldend; sampnum++)
{
firstlavg += info.lsound[sampnum];
firstravg += info.rsound[sampnum];
}
firstlavg /= firstfieldend;
firstravg /= firstfieldend;
// then compute the standard deviation over the first field
uint32_t firstldev = 0;
uint32_t firstrdev = 0;
for (uint32_t sampnum = 0; sampnum < firstfieldend; sampnum++)
{
firstldev += (info.lsound[sampnum] - firstlavg) * (info.lsound[sampnum] - firstlavg);
firstrdev += (info.rsound[sampnum] - firstravg) * (info.rsound[sampnum] - firstravg);
}
firstldev = sqrt(double(firstldev) / firstfieldend);
firstrdev = sqrt(double(firstrdev) / firstfieldend);
// scan forward through the samples, counting consecutive samples more than
// SIGNAL_DEVIATIONS standard deviations away from silence
uint32_t lcount = 0;
uint32_t rcount = 0;
uint32_t sampnum = 0;
for (sampnum = 0; sampnum < soundend; sampnum++)
{
// left speaker
if (info.lsound[sampnum] > firstlavg + SIGNAL_DEVIATIONS * firstldev)
lcount++;
else
lcount = 0;
// right speaker
if (info.rsound[sampnum] > firstravg + SIGNAL_DEVIATIONS * firstrdev)
rcount++;
else
rcount = 0;
// stop if we find enough
if (lcount > MINIMUM_SIGNAL_COUNT || rcount > MINIMUM_SIGNAL_COUNT)
break;
}
// if we didn't find any, return failure
if (sampnum >= soundend)
{
if (!report_best_field)
printf("Field %5d: Unable to find edge\n", fieldnum);
return false;
}
// scan backwards to find the start of the signal
for ( ; sampnum > 0; sampnum--)
if (info.lsound[sampnum - 1] < firstlavg + SIGNAL_START_DEVIATIONS * firstldev ||
info.rsound[sampnum - 1] < firstravg + SIGNAL_START_DEVIATIONS * firstrdev)
break;
// if we're to report the best field, figure out which field we are in
if (report_best_field)
{
int32_t curfield;
for (curfield = 0; curfield < fields_to_read - 1; curfield++)
if (sampnum < fieldstart[curfield + 1])
break;
printf("Field %5d: Edge found at offset %d (frame %.1f)\n", firstfield + curfield, sampnum - fieldstart[curfield], (double)(firstfield + curfield) * 0.5);
}
// otherwise, compute the delta from the provided field number
else
{
printf("Field %5d: Edge at offset %d from expected (found at %d, expected %d)\n", fieldnum, sampnum - targetsoundstart, sampnum, targetsoundstart);
delta = sampnum - targetsoundstart;
}
return true;
}
//-------------------------------------------------
// generate_one_frame - generate a single
// resampled frame
//-------------------------------------------------
void chd_resample_compressor::generate_one_frame(uint8_t *dest, uint32_t datasize, uint32_t fieldnum)
{
// determine the first field needed to cover this range of samples
uint32_t srcbegin = field_to_sample_number(m_info, fieldnum);
int64_t dstbegin = (int64_t(srcbegin) << 24) + m_ioffset + m_islope * fieldnum;
uint32_t dstbeginoffset;
int32_t dstbeginfield;
if (dstbegin >= 0)
dstbeginfield = sample_number_to_field(m_info, dstbegin >> 24, dstbeginoffset);
else
{
dstbeginfield = -1 - sample_number_to_field(m_info, -dstbegin >> 24, dstbeginoffset);
dstbeginoffset = (field_to_sample_number(m_info, -dstbeginfield) - field_to_sample_number(m_info, -dstbeginfield - 1)) - dstbeginoffset;
}
// determine the last field needed to cover this range of samples
uint32_t srcend = field_to_sample_number(m_info, fieldnum + 1);
int64_t dstend = (int64_t(srcend) << 24) + m_ioffset + m_islope * (fieldnum + 1);
uint32_t dstendoffset;
int32_t dstendfield;
if (dstend >= 0)
dstendfield = sample_number_to_field(m_info, dstend >> 24, dstendoffset);
else
{
dstendfield = -1 - -sample_number_to_field(m_info, -dstend >> 24, dstendoffset);
dstendoffset = (field_to_sample_number(m_info, -dstendfield) - field_to_sample_number(m_info, -dstendfield - 1)) - dstendoffset;
}
/*
printf("%5d: start=%10d (%5d.%03d) end=%10d (%5d.%03d)\n",
fieldnum,
(int32_t)(dstbegin >> 24), dstbeginfield, dstbeginoffset,
(int32_t)(dstend >> 24), dstendfield, dstendoffset);
*/
// read all samples required into the end of the sound buffers
uint32_t dstoffset = srcend - srcbegin;
for (int32_t dstfield = dstbeginfield; dstfield <= dstendfield; dstfield++)
{
if (dstfield >= 0)
read_chd(m_source, dstfield, m_info, dstoffset);
else
{
m_info.samples = field_to_sample_number(m_info, -dstfield) - field_to_sample_number(m_info, -dstfield - 1);
memset(&m_info.lsound[dstoffset], 0, m_info.samples * sizeof(m_info.lsound[0]));
memset(&m_info.rsound[dstoffset], 0, m_info.samples * sizeof(m_info.rsound[0]));
}
dstoffset += m_info.samples;
}
// resample the destination samples to the source
dstoffset = srcend - srcbegin;
int64_t dstpos = dstbegin;
int64_t dststep = (dstend - dstbegin) / int64_t(srcend - srcbegin);
for (uint32_t srcoffset = 0; srcoffset < srcend - srcbegin; srcoffset++)
{
m_info.lsound[srcoffset] = m_info.lsound[(int)(dstoffset + dstbeginoffset + (dstpos >> 24) - (dstbegin >> 24))];
m_info.rsound[srcoffset] = m_info.rsound[(int)(dstoffset + dstbeginoffset + (dstpos >> 24) - (dstbegin >> 24))];
dstpos += dststep;
}
// read the original frame, pointing the sound buffer past where we've calculated
read_chd(m_source, fieldnum, m_info, srcend - srcbegin);
// assemble the final frame
std::vector<uint8_t> buffer;
int16_t *sampledata[2] = { &m_info.lsound[0], &m_info.rsound[0] };
avhuff_encoder::assemble_data(buffer, m_info.bitmap, m_info.channels, m_info.samples, sampledata);
memcpy(dest, &buffer[0], std::min(buffer.size(), size_t(datasize)));
if (buffer.size() < datasize)
memset(&dest[buffer.size()], 0, datasize - buffer.size());
}
//-------------------------------------------------
// usage - display program usage
//-------------------------------------------------
static int usage(void)
{
fprintf(stderr, "Usage: \n");
fprintf(stderr, " ldresample source.chd\n");
fprintf(stderr, " ldresample source.chd output.chd offset [slope]\n");
fprintf(stderr, "\n");
fprintf(stderr, "Where offset and slope make a linear equation f(x) which\n");
fprintf(stderr, "describes the sample offset from the source as a function\n");
fprintf(stderr, "of field number.\n");
return 1;
}
//-------------------------------------------------
// main - main entry point
//-------------------------------------------------
int main(int argc, char *argv[])
{
// verify arguments
if (argc < 2)
return usage();
const char *srcfilename = argv[1];
const char *dstfilename = (argc < 3) ? nullptr : argv[2];
double offset = (argc < 4) ? 0.0 : atof(argv[3]);
double slope = (argc < 5) ? 1.0 : atof(argv[4]);
// print basic information
printf("Input file: %s\n", srcfilename);
if (dstfilename != nullptr)
{
printf("Output file: %s\n", dstfilename);
printf("Offset: %f\n", offset);
printf("Slope: %f\n", slope);
}
// open the source file
chd_file srcfile;
movie_info info;
chd_error err = open_chd(srcfile, srcfilename, info);
if (err != CHDERR_NONE)
{
fprintf(stderr, "Unable to open file '%s'\n", srcfilename);
return 1;
}
// output some basics
printf("Video dimensions: %dx%d\n", info.width, info.height);
printf("Video frame rate: %.2fHz\n", info.framerate);
printf("Sample rate: %dHz\n", info.samplerate);
printf("Total fields: %d\n", info.numfields);
// if we don't have a destination file, scan for edges
if (dstfilename == nullptr)
{
for (uint32_t fieldnum = 60; fieldnum < info.numfields - 60; fieldnum += 30)
{
fprintf(stderr, "Field %5d\r", fieldnum);
int32_t delta;
find_edge_near_field(srcfile, fieldnum, info, true, delta);
}
}
// otherwise, resample the source to the destination
else
{
// open the destination file
chd_resample_compressor dstfile(srcfile, info, int64_t(offset * 65536.0 * 256.0), int64_t(slope * 65536.0 * 256.0));
err = create_chd(dstfile, dstfilename, srcfile, info);
if (!dstfile.opened())
{
fprintf(stderr, "Unable to create file '%s'\n", dstfilename);
return 1;
}
// loop over all the fields in the source file
double progress, ratio;
osd_ticks_t last_update = 0;
while (dstfile.compress_continue(progress, ratio) == CHDERR_COMPRESSING)
if (osd_ticks() - last_update > osd_ticks_per_second() / 4)
{
last_update = osd_ticks();
printf("Processing, %.1f%% complete....\r", progress * 100.0);
}
}
return 0;
}