mirror of
https://gitlab.freedesktop.org/gstreamer/gstreamer.git
synced 2024-11-26 19:51:11 +00:00
349 lines
11 KiB
C
349 lines
11 KiB
C
/* GStreamer LC3 Bluetooth LE audio decoder
|
|
* Copyright (C) 2023 Asymptotic Inc. <taruntej@asymptotic.io>
|
|
*
|
|
* This library is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU Library General Public
|
|
* License as published by the Free Software Foundation; either
|
|
* version 2 of the License, or (at your option) any later version.
|
|
*
|
|
* This library is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
* Library General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU Library General Public
|
|
* License along with this library; if not, write to the
|
|
* Free Software Foundation, Inc., 51 Franklin Street, Suite 500,
|
|
* Boston, MA 02110-1335, USA.
|
|
*/
|
|
|
|
/**
|
|
* SECTION:element-lc3dec
|
|
*
|
|
* The lc3dec decodes LC3 data into raw audio.
|
|
*
|
|
* ## Example pipeline
|
|
* |[
|
|
* gst-launch-1.0 -v filesrc location=encoded.lc3 blocksize=200 ! \
|
|
* audio/x-lc3,frame-bytes=100,frame-duration-us=10000,channels=2,rate=48000,channel-mask=\(bitmask\)0x00000000000000003 !\
|
|
* lc3dec ! wavenc ! filesink location=decoded.wav
|
|
* ]|
|
|
*
|
|
* Decodes the LC3 frames each with 100 bytes of size, converts it to raw audio and saves into a .wav file
|
|
*
|
|
* Since: 1.24
|
|
*/
|
|
|
|
#ifdef HAVE_CONFIG_H
|
|
#include "config.h"
|
|
#endif
|
|
|
|
#include <gst/gst.h>
|
|
#include <gst/audio/gstaudiodecoder.h>
|
|
|
|
#include "gstlc3common.h"
|
|
#include "gstlc3dec.h"
|
|
|
|
GST_DEBUG_CATEGORY_STATIC (gst_lc3_dec_debug_category);
|
|
#define GST_CAT_DEFAULT gst_lc3_dec_debug_category
|
|
|
|
#define parent_class gst_lc3_dec_parent_class
|
|
G_DEFINE_TYPE (GstLc3Dec, gst_lc3_dec, GST_TYPE_AUDIO_DECODER);
|
|
GST_ELEMENT_REGISTER_DEFINE (lc3dec, "lc3dec", GST_RANK_NONE, GST_TYPE_LC3_DEC);
|
|
|
|
/* prototypes */
|
|
static gboolean gst_lc3_dec_start (GstAudioDecoder * decoder);
|
|
static gboolean gst_lc3_dec_stop (GstAudioDecoder * decoder);
|
|
static gboolean gst_lc3_dec_set_format (GstAudioDecoder * decoder,
|
|
GstCaps * caps);
|
|
static GstFlowReturn gst_lc3_dec_handle_frame (GstAudioDecoder * decoder,
|
|
GstBuffer * buffer);
|
|
|
|
/* pad templates */
|
|
static GstStaticPadTemplate gst_lc3_dec_src_template =
|
|
GST_STATIC_PAD_TEMPLATE ("src",
|
|
GST_PAD_SRC,
|
|
GST_PAD_ALWAYS,
|
|
GST_STATIC_CAPS ("audio/x-raw, "
|
|
"format = " FORMAT ", layout=interleaved, "
|
|
"rate = { " SAMPLE_RATES " }, channels = [1,MAX]")
|
|
);
|
|
|
|
static GstStaticPadTemplate gst_lc3_dec_sink_template =
|
|
GST_STATIC_PAD_TEMPLATE ("sink",
|
|
GST_PAD_SINK,
|
|
GST_PAD_ALWAYS,
|
|
GST_STATIC_CAPS ("audio/x-lc3, rate = { " SAMPLE_RATES " }, "
|
|
"channels = [1,MAX],"
|
|
"frame-bytes = (int) [" FRAME_BYTES_RANGE "], "
|
|
"frame-duration-us = (int) { " FRAME_DURATIONS " }, "
|
|
"framed=(boolean) true")
|
|
);
|
|
|
|
/* class initialization */
|
|
static void
|
|
gst_lc3_dec_class_init (GstLc3DecClass * klass)
|
|
{
|
|
GstAudioDecoderClass *audio_decoder_class = GST_AUDIO_DECODER_CLASS (klass);
|
|
|
|
gst_element_class_add_static_pad_template (GST_ELEMENT_CLASS (klass),
|
|
&gst_lc3_dec_src_template);
|
|
gst_element_class_add_static_pad_template (GST_ELEMENT_CLASS (klass),
|
|
&gst_lc3_dec_sink_template);
|
|
|
|
gst_element_class_set_static_metadata (GST_ELEMENT_CLASS (klass),
|
|
"LC3 Bluetooth Audio decoder", "Codec/Decoder/Audio",
|
|
"Decodes an LC3 Audio stream to raw audio",
|
|
"Taruntej Kanakamalla <taruntej@asymptotic.io>");
|
|
|
|
GST_DEBUG_CATEGORY_INIT (gst_lc3_dec_debug_category, "lc3dec", 0,
|
|
"debug category for lc3dec element");
|
|
|
|
audio_decoder_class->start = GST_DEBUG_FUNCPTR (gst_lc3_dec_start);
|
|
audio_decoder_class->stop = GST_DEBUG_FUNCPTR (gst_lc3_dec_stop);
|
|
audio_decoder_class->set_format = GST_DEBUG_FUNCPTR (gst_lc3_dec_set_format);
|
|
audio_decoder_class->handle_frame =
|
|
GST_DEBUG_FUNCPTR (gst_lc3_dec_handle_frame);
|
|
}
|
|
|
|
static void
|
|
gst_lc3_dec_init (GstLc3Dec * lc3_dec)
|
|
{
|
|
}
|
|
|
|
static gboolean
|
|
gst_lc3_dec_start (GstAudioDecoder * decoder)
|
|
{
|
|
/* let the baseclass convert the segment data
|
|
* from 'bytes' to 'time' format
|
|
*/
|
|
gst_audio_decoder_set_estimate_rate (decoder, TRUE);
|
|
|
|
/* Inform the base class that the LC3 lib can do PLC */
|
|
gst_audio_decoder_set_plc_aware (decoder, TRUE);
|
|
|
|
return TRUE;
|
|
}
|
|
|
|
static gboolean
|
|
gst_lc3_dec_stop (GstAudioDecoder * decoder)
|
|
{
|
|
GstLc3Dec *lc3_dec = GST_LC3_DEC (decoder);
|
|
|
|
if (lc3_dec->dec_ch != NULL) {
|
|
for (int ich = 0; ich < lc3_dec->channels; ich++) {
|
|
g_free (lc3_dec->dec_ch[ich]);
|
|
lc3_dec->dec_ch[ich] = NULL;
|
|
}
|
|
|
|
g_free (lc3_dec->dec_ch);
|
|
lc3_dec->dec_ch = NULL;
|
|
}
|
|
|
|
return TRUE;
|
|
}
|
|
|
|
static gboolean
|
|
gst_lc3_dec_set_format (GstAudioDecoder * decoder, GstCaps * caps)
|
|
{
|
|
GstLc3Dec *lc3_dec = GST_LC3_DEC (decoder);
|
|
GstAudioInfo info;
|
|
GstStructure *s;
|
|
GstAudioChannelPosition pos[64] = { GST_AUDIO_CHANNEL_POSITION_INVALID, };
|
|
gint in_ch, in_rate;
|
|
guint64 in_chmsk = 0;
|
|
GstClockTime latency;
|
|
|
|
GST_DEBUG_OBJECT (lc3_dec, "set_format");
|
|
GST_DEBUG_OBJECT (lc3_dec, "input caps %" GST_PTR_FORMAT, caps);
|
|
|
|
s = gst_caps_get_structure (caps, 0);
|
|
if (!gst_structure_get_int (s, "frame-duration-us",
|
|
&lc3_dec->frame_duration_us)) {
|
|
GST_ERROR_OBJECT (lc3_dec,
|
|
"sink caps does not contain 'frame-duration-us'");
|
|
return FALSE;
|
|
}
|
|
|
|
if (!gst_structure_get_int (s, "frame-bytes", &lc3_dec->frame_bytes)) {
|
|
GST_ERROR_OBJECT (lc3_dec, "sink caps does not contain 'frame-bytes'");
|
|
return FALSE;
|
|
}
|
|
/* use rate and channel from input caps to create filter caps */
|
|
gst_structure_get_int (s, "rate", &in_rate);
|
|
gst_structure_get_int (s, "channels", &in_ch);
|
|
if (!gst_structure_get (s, "channel-mask", GST_TYPE_BITMASK, &in_chmsk, NULL)) {
|
|
GST_INFO_OBJECT (lc3_dec,
|
|
"channel-mask not present in the sink caps, getting fallback mask");
|
|
in_chmsk = gst_audio_channel_get_fallback_mask (in_ch);
|
|
}
|
|
s = NULL;
|
|
|
|
gst_audio_channel_positions_from_mask (in_ch, in_chmsk, pos);
|
|
gst_audio_info_set_format (&info, GST_AUDIO_FORMAT_S16LE, in_rate, in_ch,
|
|
pos);
|
|
|
|
/* get rate, format, channels from the output caps */
|
|
lc3_dec->rate = GST_AUDIO_INFO_RATE (&info);
|
|
lc3_dec->channels = GST_AUDIO_INFO_CHANNELS (&info);
|
|
|
|
switch (GST_AUDIO_INFO_FORMAT (&info)) {
|
|
case GST_AUDIO_FORMAT_S16LE:
|
|
lc3_dec->format = LC3_PCM_FORMAT_S16;
|
|
break;
|
|
case GST_AUDIO_FORMAT_S24LE:
|
|
lc3_dec->format = LC3_PCM_FORMAT_S24_3LE;
|
|
break;
|
|
case GST_AUDIO_FORMAT_F32:
|
|
lc3_dec->format = LC3_PCM_FORMAT_FLOAT;
|
|
break;
|
|
case GST_AUDIO_FORMAT_S24_32LE:
|
|
default:
|
|
lc3_dec->format = LC3_PCM_FORMAT_S24;
|
|
break;
|
|
}
|
|
|
|
GST_INFO_OBJECT (lc3_dec, "lc3dec params "
|
|
"rate: %" G_GINT32_FORMAT ", channels: %" G_GINT32_FORMAT
|
|
", lc3_pcm_format = %" G_GINT32_FORMAT " frame len: %" G_GINT32_FORMAT
|
|
", frame_duration " "%" G_GINT32_FORMAT, lc3_dec->rate, lc3_dec->channels,
|
|
lc3_dec->format, lc3_dec->frame_bytes, lc3_dec->frame_duration_us);
|
|
|
|
lc3_dec->frame_samples =
|
|
lc3_frame_samples (lc3_dec->frame_duration_us, lc3_dec->rate);
|
|
lc3_dec->bpf = GST_AUDIO_INFO_BPF (&info);
|
|
|
|
latency =
|
|
gst_util_uint64_scale_int (lc3_dec->frame_bytes, GST_SECOND,
|
|
lc3_dec->rate);
|
|
gst_audio_decoder_set_latency (decoder, latency, latency);
|
|
|
|
/* Setup and Init decoder handle */
|
|
if (lc3_dec->dec_ch != NULL) {
|
|
for (int ich = 0; ich < lc3_dec->channels; ich++) {
|
|
g_free (lc3_dec->dec_ch[ich]);
|
|
lc3_dec->dec_ch[ich] = NULL;
|
|
}
|
|
g_free (lc3_dec->dec_ch);
|
|
lc3_dec->dec_ch = NULL;
|
|
}
|
|
|
|
lc3_dec->dec_ch = g_new0 (lc3_decoder_t, lc3_dec->channels);
|
|
|
|
for (guint8 i = 0; i < lc3_dec->channels; i++) {
|
|
/* The decoder can resample for us. But we leave the resampling to before decoding
|
|
* explicitly for now. So pass the same sample rate for sr_hz and sr_pcm_hz
|
|
*/
|
|
lc3_dec->dec_ch[i] =
|
|
lc3_setup_decoder (lc3_dec->frame_duration_us, lc3_dec->rate,
|
|
lc3_dec->rate, g_malloc (lc3_decoder_size (lc3_dec->frame_duration_us,
|
|
lc3_dec->rate)));
|
|
|
|
if (lc3_dec->dec_ch[i] == NULL) {
|
|
GST_ERROR_OBJECT (lc3_dec,
|
|
"Failed to create decoder handle for channel %" G_GUINT32_FORMAT, i);
|
|
return FALSE;
|
|
}
|
|
}
|
|
|
|
gst_audio_decoder_set_output_format (decoder, &info);
|
|
return TRUE;
|
|
}
|
|
|
|
static GstFlowReturn
|
|
gst_lc3_dec_handle_frame (GstAudioDecoder * decoder, GstBuffer * inbuf)
|
|
{
|
|
GstLc3Dec *lc3_dec = GST_LC3_DEC (decoder);
|
|
GstBuffer *outbuf = NULL;
|
|
GstMapInfo out_map;
|
|
GstMapInfo in_map;
|
|
gssize output_size;
|
|
GstAudioClippingMeta *audio_meta;
|
|
gboolean do_plc = gst_audio_decoder_get_plc (decoder) &&
|
|
gst_audio_decoder_get_plc_aware (decoder);
|
|
|
|
/* no fancy draining */
|
|
if (G_UNLIKELY (inbuf == NULL))
|
|
return GST_FLOW_OK;
|
|
|
|
gst_buffer_map (inbuf, &in_map, GST_MAP_READ);
|
|
|
|
if (G_UNLIKELY (in_map.size == 0 && !do_plc)) {
|
|
GST_ERROR_OBJECT (lc3_dec,
|
|
"PLC handled by the base class, should not get a zero sized buffer");
|
|
return GST_FLOW_ERROR;
|
|
}
|
|
|
|
GST_LOG_OBJECT (lc3_dec, "received %lu bytes ", in_map.size);
|
|
|
|
/* we expect exactly one frame each time */
|
|
if (G_UNLIKELY (in_map.size == 0 && !do_plc) &&
|
|
(in_map.size != (lc3_dec->frame_bytes * lc3_dec->channels)))
|
|
goto mixed_frames;
|
|
|
|
output_size = lc3_dec->frame_samples * lc3_dec->bpf;
|
|
GST_LOG_OBJECT (lc3_dec, "allocating %lu bytes to output buffer",
|
|
output_size);
|
|
outbuf = gst_audio_decoder_allocate_output_buffer (decoder, output_size);
|
|
|
|
if (outbuf == NULL)
|
|
goto no_buffer;
|
|
|
|
gst_buffer_map (outbuf, &out_map, GST_MAP_WRITE);
|
|
|
|
for (guint c = 0; c < lc3_dec->channels; c++) {
|
|
gint ret = 0;
|
|
void *in = in_map.data ? in_map.data + (c * lc3_dec->frame_bytes) : NULL;
|
|
ret =
|
|
lc3_decode (lc3_dec->dec_ch[c], in, lc3_dec->frame_bytes,
|
|
lc3_dec->format, out_map.data + (c * lc3_dec->bpf / lc3_dec->channels),
|
|
lc3_dec->channels);
|
|
|
|
if (ret < 0) {
|
|
GST_ERROR_OBJECT (lc3_dec,
|
|
"Failed to decode frame for buffer %" GST_PTR_FORMAT, inbuf);
|
|
return GST_FLOW_ERROR;
|
|
} else if (ret == 1) {
|
|
GST_INFO_OBJECT (lc3_dec, "PLC operated for channel: %d", c + 1);
|
|
}
|
|
}
|
|
|
|
audio_meta = gst_buffer_get_audio_clipping_meta (inbuf);
|
|
if (audio_meta) {
|
|
switch (audio_meta->format) {
|
|
case GST_FORMAT_DEFAULT:
|
|
{
|
|
output_size =
|
|
output_size - (audio_meta->start * lc3_dec->bpf) -
|
|
(audio_meta->end * lc3_dec->bpf);
|
|
gst_buffer_resize (outbuf, (audio_meta->start * lc3_dec->bpf),
|
|
output_size);
|
|
}
|
|
break;
|
|
default:
|
|
GST_WARNING_OBJECT (lc3_dec, "audio meta format: %d not handled",
|
|
audio_meta->format);
|
|
break;
|
|
}
|
|
}
|
|
|
|
gst_buffer_unmap (outbuf, &out_map);
|
|
gst_buffer_unmap (inbuf, &in_map);
|
|
|
|
return gst_audio_decoder_finish_frame (decoder, outbuf, 1);
|
|
|
|
/* ERRORS */
|
|
mixed_frames:
|
|
{
|
|
GST_WARNING_OBJECT (lc3_dec,
|
|
"inconsistent input data/frames, Need to be %"
|
|
G_GINT32_FORMAT " bytes", lc3_dec->frame_bytes * lc3_dec->channels);
|
|
return GST_FLOW_ERROR;
|
|
}
|
|
|
|
no_buffer:
|
|
{
|
|
GST_ERROR_OBJECT (lc3_dec, "could not allocate output buffer");
|
|
return GST_FLOW_ERROR;
|
|
}
|
|
}
|