Skip to content

Audio: MFCC: Use the MFCC module as compress PCM encoder with discontinuous stream #10814

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
singalsu wants to merge 4 commits into
base: main
Choose a base branch
from
Open

Audio: MFCC: Use the MFCC module as compress PCM encoder with discontinuous stream #10814

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
85364a2
audio: mfcc: switch to source/sink API, int32 output, and DTX
singalsu May 22, 2026
7c6e487
base_fw: advertise BESPOKE codec for MFCC compress capture
singalsu May 26, 2026
0e7e06f
audio: mfcc: update decode tools and add Python compress scripts
singalsu May 26, 2026
b491cf1
tools: topology: add MFCC compress capture for jack and DMIC
singalsu May 26, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions src/audio/base_fw.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -100,6 +100,10 @@ static void get_codec_info(struct sof_tlv **tuple)
codec_info.items[codec_info.count++] =
SET_CODEC_INFO_ITEM(SND_AUDIOCODEC_VORBIS, SOF_IPC_STREAM_PLAYBACK);
#endif
#ifdef CONFIG_COMP_MFCC
codec_info.items[codec_info.count++] =
SET_CODEC_INFO_ITEM(SND_AUDIOCODEC_BESPOKE, SOF_IPC_STREAM_CAPTURE);
#endif

if (!codec_info.count)
return;
Expand Down
164 changes: 131 additions & 33 deletions src/audio/mfcc/mfcc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,6 +12,8 @@
#include <sof/audio/format.h>
#include <sof/audio/pipeline.h>
#include <sof/audio/ipc-config.h>
#include <module/audio/source_api.h>
#include <module/audio/sink_api.h>
#include <sof/common.h>
#include <rtos/panic.h>
#include <sof/ipc/msg.h>
Expand All @@ -36,29 +38,39 @@ LOG_MODULE_REGISTER(mfcc, CONFIG_SOF_LOG_LEVEL);

SOF_DEFINE_REG_UUID(mfcc);

__cold_rodata const struct mfcc_func_map mfcc_fm[] = {
/** \brief Source/sink API based source copy function map. */
struct mfcc_source_func_map {
uint8_t source;
mfcc_source_func func;
};

__cold_rodata static const struct mfcc_source_func_map mfcc_sfm[] = {
#if CONFIG_FORMAT_S16LE
{SOF_IPC_FRAME_S16_LE, mfcc_s16_default},
#endif /* CONFIG_FORMAT_S16LE */
{SOF_IPC_FRAME_S16_LE, mfcc_source_copy_s16},
#endif
#if CONFIG_FORMAT_S24LE
{SOF_IPC_FRAME_S24_4LE, mfcc_s24_default},
#endif /* CONFIG_FORMAT_S24LE */
{SOF_IPC_FRAME_S24_4LE, mfcc_source_copy_s24},
#endif
#if CONFIG_FORMAT_S32LE
{SOF_IPC_FRAME_S32_LE, mfcc_s32_default},
#endif /* CONFIG_FORMAT_S32LE */
{SOF_IPC_FRAME_S32_LE, mfcc_source_copy_s32},
#endif
};

static mfcc_func mfcc_find_func(enum sof_ipc_frame source_format,
enum sof_ipc_frame sink_format,
const struct mfcc_func_map *map,
int n)
/**
* \brief Look up the source copy function for a given input format.
*
* \param[in] source_format SOF IPC frame format of the input stream.
*
* \return Pointer to the matching mfcc_source_func, or NULL if the
* format is not supported in this build.
*/
static mfcc_source_func mfcc_find_source_func(enum sof_ipc_frame source_format)
{
int i;

/* Find suitable processing function from map. */
for (i = 0; i < n; i++) {
if (source_format == map[i].source)
return map[i].func;
for (i = 0; i < ARRAY_SIZE(mfcc_sfm); i++) {
if (source_format == mfcc_sfm[i].source)
return mfcc_sfm[i].func;
}

return NULL;
Expand All @@ -68,6 +80,16 @@ static mfcc_func mfcc_find_func(enum sof_ipc_frame source_format,
* End of MFCC setup code. Next the standard component methods.
*/

/**
* \brief MFCC module init callback.
*
* Allocates the component private data and the configuration blob
* handler used to receive the MFCC configuration from user space.
*
* \param[in,out] mod Processing module being initialized.
*
* \return 0 on success, -ENOMEM on allocation failure.
*/
static int mfcc_init(struct processing_module *mod)
{
struct module_data *md = &mod->priv;
Expand All @@ -92,6 +114,16 @@ static int mfcc_init(struct processing_module *mod)
return 0;
}

/**
* \brief MFCC module free callback.
*
* Releases the IPC notification message, configuration blob handler,
* runtime buffers and the private data structure.
*
* \param[in,out] mod Processing module being freed.
*
* \return 0 on success.
*/
static int mfcc_free(struct processing_module *mod)
{
struct mfcc_comp_data *cd = module_get_private_data(mod);
Expand All @@ -106,27 +138,74 @@ static int mfcc_free(struct processing_module *mod)
}


/**
* \brief Source/sink API based process function for MFCC.
*
* Reads input audio from sof_source, runs the STFT/Mel/DCT stage, and
* delegates output formatting and commit handling to mfcc_common.c.
*
* \param[in,out] mod Processing module.
* \param[in,out] sources Source array; only sources[0] is used.
* \param[in] num_of_sources Number of sources (must be 1).
* \param[in,out] sinks Sink array; only sinks[0] is used.
* \param[in] num_of_sinks Number of sinks (must be 1).
*
* \return 0 on success, -ENODATA if no input frames are available,
* or a negative error code from the STFT or output stages.
*/
static int mfcc_process(struct processing_module *mod,
struct input_stream_buffer *input_buffers, int num_input_buffers,
struct output_stream_buffer *output_buffers, int num_output_buffers)
struct sof_source **sources, int num_of_sources,
struct sof_sink **sinks, int num_of_sinks)
{
struct mfcc_comp_data *cd = module_get_private_data(mod);
struct audio_stream *source = input_buffers->data;
struct audio_stream *sink = output_buffers->data;
int frames = input_buffers->size;
struct comp_dev *dev = mod->dev;
struct mfcc_state *state = &cd->state;
bool pending;
size_t source_avail;
int frames;
int num_ceps;

comp_dbg(mod->dev, "start");
comp_dbg(dev, "start");

frames = MIN(frames, cd->max_frames);
cd->mfcc_func(mod, input_buffers, output_buffers, frames);
/* In compress mode, retry pending output first and avoid producing
* new frames until previous frame has been committed.
*/
pending = state->header_pending || state->out_remain > 0;
if (cd->config->compress_output && pending)
return mfcc_process_output(mod, cd, sources, sinks, 0, 0);

/* TODO: use module_update_buffer_position() from #6194 */
input_buffers->consumed += audio_stream_frame_bytes(source) * frames;
output_buffers->size += audio_stream_frame_bytes(sink) * frames;
comp_dbg(mod->dev, "done");
return 0;
source_avail = source_get_data_frames_available(sources[0]);
frames = MIN(source_avail, cd->max_frames);
if (frames == 0)
return -ENODATA;

/* Copy input audio from source to MFCC internal circular buffer */
cd->source_func(sources[0], &state->buf, &state->emph, frames, state->source_channel);
Comment thread
singalsu marked this conversation as resolved.

/* Run STFT and Mel/DCT processing */
num_ceps = mfcc_stft_process(mod, cd);
if (num_ceps < 0)
return num_ceps;

return mfcc_process_output(mod, cd, sources, sinks, num_ceps, frames);
}

/**
* \brief MFCC module prepare callback.
*
* Validates the source/sink connection, reads the configuration blob,
* initializes the MFCC processing state via mfcc_setup(), selects the
* input copy function for the source frame format, and prepares the
* VAD switch control notification when enabled.
*
* \param[in,out] mod Processing module being prepared.
* \param[in,out] sources Source array; only sources[0] is used.
* \param[in] num_of_sources Number of sources (must be 1).
* \param[in,out] sinks Sink array; only sinks[0] is used.
* \param[in] num_of_sinks Number of sinks (must be 1).
*
* \return 0 on success or a negative error code.
*/
static int mfcc_prepare(struct processing_module *mod,
struct sof_source **sources, int num_of_sources,
struct sof_sink **sinks, int num_of_sinks)
Expand Down Expand Up @@ -172,13 +251,21 @@ static int mfcc_prepare(struct processing_module *mod,
return -EINVAL;
}

cd->mfcc_func = mfcc_find_func(source_format, sink_format, mfcc_fm, ARRAY_SIZE(mfcc_fm));
if (!cd->mfcc_func) {
comp_err(dev, "No proc func");
cd->source_func = mfcc_find_source_func(source_format);
if (!cd->source_func) {
comp_err(dev, "No source func");
mfcc_free_buffers(mod);
return -EINVAL;
}

cd->source_format = source_format;

if (cd->config->compress_output)
comp_info(dev, "compress PCM output mode enabled");

Comment on lines +261 to +265
Copy link
Copy Markdown
Collaborator Author

@singalsu singalsu May 28, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks — but the sink format is intentionally unconstrained here. mfcc_output_legacy() computes commit_bytes = sink_get_frame_bytes(sink) * frames, zero-fills that period, then copies the header + int32 payload byte-wise into the period, carrying any leftover via state->out_remain. The commit size always matches what the sink expects, so S16_LE / S24_4LE / S32_LE sinks all produce correctly-sized commits — no truncation or buffer overrun. The bench topologies that wire S16_LE on the MFCC sink rely on this behavior (the host decodes the bytes as an MFCC blob, not as PCM). Rejecting non-S32 sinks would break those bench flows. The only thing being conveyed through the sink is opaque bytes; there is no PCM-format contract on the MFCC output.

if (cd->config->enable_dtx && !cd->config->compress_output)
comp_warn(dev, "enable_dtx ignored in normal PCM mode, only applies to compress");

/* Initialize VAD switch control notification if enabled */
if (cd->config->enable_vad && cd->config->update_controls) {
if (!cd->msg) {
Expand All @@ -194,6 +281,17 @@ static int mfcc_prepare(struct processing_module *mod,
return 0;
}

/**
* \brief MFCC module reset callback.
*
* Frees runtime buffers (NULLing their pointers) so a subsequent
* mfcc_prepare() can re-allocate cleanly. The configuration blob
* handler and IPC notification message are preserved.
*
* \param[in,out] mod Processing module being reset.
*
* \return 0 on success.
*/
static int mfcc_reset(struct processing_module *mod)
{
struct mfcc_comp_data *cd = module_get_private_data(mod);
Expand All @@ -206,7 +304,7 @@ static int mfcc_reset(struct processing_module *mod)
mfcc_free_buffers(mod);

/* Reset to similar state as init() */
cd->mfcc_func = NULL;
cd->source_func = NULL;
return 0;
}

Expand All @@ -215,7 +313,7 @@ static const struct module_interface mfcc_interface = {
.free = mfcc_free,
.set_configuration = mfcc_set_config,
.get_configuration = mfcc_get_config,
.process_audio_stream = mfcc_process,
.process = mfcc_process,
.prepare = mfcc_prepare,
.reset = mfcc_reset,
};
Expand Down
Loading
Loading