tempo scaling audio filter – an implementation of WSOLA algorithm More...

#include <float.h>
#include "libavcodec/avfft.h"
#include "libavutil/avassert.h"
#include "libavutil/avstring.h"
#include "libavutil/channel_layout.h"
#include "libavutil/eval.h"
#include "libavutil/opt.h"
#include "libavutil/samplefmt.h"
#include "avfilter.h"
#include "audio.h"
#include "internal.h"

Data Structures
struct	AudioFragment
	A fragment of audio waveform. More...

struct	ATempoContext
	Filter state machine. More...

Macros
#define	OFFSET(x) offsetof(ATempoContext, x)

#define	RE_MALLOC_OR_FAIL(field, field_size)

#define	yae_init_xdat(scalar_type, scalar_max)
	A helper macro for initializing complex data buffer with scalar data of a given type.

#define	yae_blend(scalar_type)
	A helper macro for blending the overlap region of previous and current audio fragment.

Enumerations
enum	FilterState { YAE_LOAD_FRAGMENT, YAE_ADJUST_POSITION, YAE_RELOAD_FRAGMENT, YAE_OUTPUT_OVERLAP_ADD, YAE_FLUSH_OUTPUT }
	Filter state machine states. More...

Functions
	AVFILTER_DEFINE_CLASS (atempo)

static AudioFragment *	yae_curr_frag (ATempoContext *atempo)

static AudioFragment *	yae_prev_frag (ATempoContext *atempo)

static void	yae_clear (ATempoContext *atempo)
	Reset filter to initial state, do not deallocate existing local buffers.

static void	yae_release_buffers (ATempoContext *atempo)
	Reset filter to initial state and deallocate all buffers.

static int	yae_reset (ATempoContext *atempo, enum AVSampleFormat format, int sample_rate, int channels)
	Prepare filter for processing audio data of given format, sample rate and number of channels.

static int	yae_set_tempo (AVFilterContext ctx, const char arg_tempo)

static void	yae_downmix (ATempoContext atempo, AudioFragment frag)
	Initialize complex data buffer of a given audio fragment with down-mixed mono data of appropriate scalar type.

static int	yae_load_data (ATempoContext atempo, const uint8_t src_ref, const uint8_t src_end, int64_t stop_here)
	Populate the internal data buffer on as-needed basis.

static int	yae_load_frag (ATempoContext atempo, const uint8_t src_ref, const uint8_t src_end)
	Populate current audio fragment data buffer.

static void	yae_advance_to_next_frag (ATempoContext *atempo)
	Prepare for loading next audio fragment.

static void	yae_xcorr_via_rdft (FFTSample xcorr, RDFTContext complex_to_real, const FFTComplex xa, const FFTComplex xb, const int window)
	Calculate cross-correlation via rDFT.

static int	yae_align (AudioFragment frag, const AudioFragment prev, const int window, const int delta_max, const int drift, FFTSample correlation, RDFTContext complex_to_real)
	Calculate alignment offset for given fragment relative to the previous fragment.

static int	yae_adjust_position (ATempoContext *atempo)
	Adjust current fragment position for better alignment with previous fragment.

static int	yae_overlap_add (ATempoContext atempo, uint8_t dst_ref, uint8_t dst_end)
	Blend the overlap region of previous and current audio fragment and output the results to the given destination buffer.

static void	yae_apply (ATempoContext atempo, const uint8_t src_ref, const uint8_t src_end, uint8_t *dst_ref, uint8_t dst_end)
	Feed as much data to the filter as it is able to consume and receive as much processed data in the destination buffer as it is able to produce or store.

static int	yae_flush (ATempoContext atempo, uint8_t dst_ref, uint8_t dst_end)
	Flush any buffered data from the filter.

static av_cold int	init (AVFilterContext *ctx)

static av_cold void	uninit (AVFilterContext *ctx)

static int	query_formats (AVFilterContext *ctx)

static int	config_props (AVFilterLink *inlink)

static int	push_samples (ATempoContext atempo, AVFilterLink outlink, int n_out)

static int	filter_frame (AVFilterLink inlink, AVFrame src_buffer)

static int	request_frame (AVFilterLink *outlink)

static int	process_command (AVFilterContext ctx, const char cmd, const char arg, char res, int res_len, int flags)

Variables
static const AVOption	atempo_options []

static const AVFilterPad	atempo_inputs []

static const AVFilterPad	atempo_outputs []

AVFilter	avfilter_af_atempo

Detailed Description

tempo scaling audio filter – an implementation of WSOLA algorithm

Based on MIT licensed yaeAudioTempoFilter.h and yaeAudioFragment.h from Apprentice Video player by Pavel Koshevoy. https://sourceforge.net/projects/apprenticevideo/

An explanation of SOLA algorithm is available at http://www.surina.net/article/time-and-pitch-scaling.html

WSOLA is very similar to SOLA, only one major difference exists between these algorithms. SOLA shifts audio fragments along the output stream, where as WSOLA shifts audio fragments along the input stream.

The advantage of WSOLA algorithm is that the overlap region size is always the same, therefore the blending function is constant and can be precomputed.

Definition in file af_atempo.c.

Macro Definition Documentation

#define OFFSET ( x ) offsetof(ATempoContext, x)

Definition at line 152 of file af_atempo.c.

#define RE_MALLOC_OR_FAIL	(	field,
		field_size
	)

Value:

do {                                                        \
        av_freep(&field);                                       \
        field = av_malloc(field_size);                          \
        if (!field) {                                           \
            yae_release_buffers(atempo);                        \
            return AVERROR(ENOMEM);                             \
        }                                                       \
    } while (0)

Definition at line 238 of file af_atempo.c.

Referenced by yae_reset().

#define yae_init_xdat	(	scalar_type,
		scalar_max
	)

A helper macro for initializing complex data buffer with scalar data of a given type.

Definition at line 351 of file af_atempo.c.

Referenced by yae_downmix().

#define yae_blend ( scalar_type )

Value:

do {                                                                \
        const scalar_type *aaa = (const scalar_type *)a;                \
        const scalar_type *bbb = (const scalar_type *)b;                \
                                                                        \
        scalar_type *out     = (scalar_type *)dst;                      \
        scalar_type *out_end = (scalar_type *)dst_end;                  \
        int64_t i;                                                      \
                                                                        \
        for (i = 0; i < overlap && out < out_end;                       \
             i++, atempo->position[1]++, wa++, wb++) {                  \
            float w0 = *wa;                                             \
            float w1 = *wb;                                             \
            int j;                                                      \
                                                                        \
            for (j = 0; j < atempo->channels;                           \
                 j++, aaa++, bbb++, out++) {                            \
                float t0 = (float)*aaa;                                 \
                float t1 = (float)*bbb;                                 \
                                                                        \
                *out =                                                  \
                    frag->position[0] + i < 0 ?                         \
                    *aaa :                                              \
                    (scalar_type)(t0 * w0 + t1 * w1);                   \
            }                                                           \
        }                                                               \
        dst = (uint8_t *)out;                                           \
    } while (0)

A helper macro for blending the overlap region of previous and current audio fragment.

Definition at line 732 of file af_atempo.c.

Referenced by yae_overlap_add().

Enumeration Type Documentation

enum FilterState

Filter state machine states.

Enumerator:

YAE_LOAD_FRAGMENT
YAE_ADJUST_POSITION
YAE_RELOAD_FRAGMENT
YAE_OUTPUT_OVERLAP_ADD
YAE_FLUSH_OUTPUT

Definition at line 76 of file af_atempo.c.

Function Documentation

AVFILTER_DEFINE_CLASS ( atempo )

static AudioFragment* yae_curr_frag ( ATempoContext * atempo )

inlinestatic

Definition at line 163 of file af_atempo.c.

Referenced by yae_adjust_position(), yae_advance_to_next_frag(), yae_apply(), yae_flush(), yae_load_frag(), and yae_overlap_add().

static AudioFragment* yae_prev_frag ( ATempoContext * atempo )

inlinestatic

Definition at line 168 of file af_atempo.c.

Referenced by yae_adjust_position(), yae_advance_to_next_frag(), yae_overlap_add(), and yae_set_tempo().

static void yae_clear ( ATempoContext * atempo )

static

Reset filter to initial state, do not deallocate existing local buffers.

Definition at line 176 of file af_atempo.c.

Referenced by yae_release_buffers(), and yae_reset().

static void yae_release_buffers ( ATempoContext * atempo )

static

Reset filter to initial state and deallocate all buffers.

Definition at line 216 of file af_atempo.c.

Referenced by uninit(), and yae_reset().

static int yae_reset	(	ATempoContext *	atempo,
		enum AVSampleFormat	format,
		int	sample_rate,
		int	channels
	)

static

Prepare filter for processing audio data of given format, sample rate and number of channels.

Definition at line 252 of file af_atempo.c.

Referenced by config_props().

static int yae_set_tempo	(	AVFilterContext *	ctx,
		const char *	arg_tempo
	)

static

Definition at line 322 of file af_atempo.c.

Referenced by process_command().

static void yae_downmix	(	ATempoContext *	atempo,
		AudioFragment *	frag
	)

static

Initialize complex data buffer of a given audio fragment with down-mixed mono data of appropriate scalar type.

Definition at line 401 of file af_atempo.c.

Referenced by yae_apply(), and yae_flush().

static int yae_load_data	(	ATempoContext *	atempo,
		const uint8_t **	src_ref,
		const uint8_t *	src_end,
		int64_t	stop_here
	)

static

Populate the internal data buffer on as-needed basis.

Returns: 0 if requested data was already available or was successfully loaded, AVERROR(EAGAIN) if more input data is required.

Definition at line 429 of file af_atempo.c.

Referenced by yae_load_frag().

static int yae_load_frag	(	ATempoContext *	atempo,
		const uint8_t **	src_ref,
		const uint8_t *	src_end
	)

static

Populate current audio fragment data buffer.

Returns: 0 when the fragment is ready, AVERROR(EAGAIN) if more input data is required.

Definition at line 504 of file af_atempo.c.

Referenced by yae_apply(), and yae_flush().

static void yae_advance_to_next_frag ( ATempoContext * atempo )

static

Prepare for loading next audio fragment.

Definition at line 584 of file af_atempo.c.

Referenced by yae_apply().

static void yae_xcorr_via_rdft	(	FFTSample *	xcorr,
		RDFTContext *	complex_to_real,
		const FFTComplex *	xa,
		const FFTComplex *	xb,
		const int	window
	)

static

Calculate cross-correlation via rDFT.

Multiply two vectors of complex numbers (result of real_to_complex rDFT) and transform back via complex_to_real rDFT.

Definition at line 606 of file af_atempo.c.

Referenced by yae_align().

static int yae_align	(	AudioFragment *	frag,
		const AudioFragment *	prev,
		const int	window,
		const int	delta_max,
		const int	drift,
		FFTSample *	correlation,
		RDFTContext *	complex_to_real
	)

static

Calculate alignment offset for given fragment relative to the previous fragment.

Returns: alignment offset of current fragment relative to previous.

Definition at line 640 of file af_atempo.c.

Referenced by yae_adjust_position().

static int yae_adjust_position ( ATempoContext * atempo )

static

Adjust current fragment position for better alignment with previous fragment.

Returns: alignment correction.

Definition at line 694 of file af_atempo.c.

Referenced by yae_apply(), and yae_flush().

static int yae_overlap_add	(	ATempoContext *	atempo,
		uint8_t **	dst_ref,
		uint8_t *	dst_end
	)

static

Blend the overlap region of previous and current audio fragment and output the results to the given destination buffer.

Returns: 0 if the overlap region was completely stored in the dst buffer, AVERROR(EAGAIN) if more destination buffer space is required.

Definition at line 769 of file af_atempo.c.

Referenced by yae_apply(), and yae_flush().

static void yae_apply	(	ATempoContext *	atempo,
		const uint8_t **	src_ref,
		const uint8_t *	src_end,
		uint8_t **	dst_ref,
		uint8_t *	dst_end
	)

static

Feed as much data to the filter as it is able to consume and receive as much processed data in the destination buffer as it is able to produce or store.

Definition at line 824 of file af_atempo.c.

Referenced by filter_frame().

static int yae_flush	(	ATempoContext *	atempo,
		uint8_t **	dst_ref,
		uint8_t *	dst_end
	)

static

Flush any buffered data from the filter.

Returns: 0 if all data was completely stored in the dst buffer, AVERROR(EAGAIN) if more destination buffer space is required.

Definition at line 898 of file af_atempo.c.

Referenced by request_frame().

static av_cold int init ( AVFilterContext * ctx )

static

Definition at line 976 of file af_atempo.c.

static av_cold void uninit ( AVFilterContext * ctx )

static

Definition at line 984 of file af_atempo.c.

static int query_formats ( AVFilterContext * ctx )

static

Definition at line 990 of file af_atempo.c.

static int config_props ( AVFilterLink * inlink )

static

Definition at line 1031 of file af_atempo.c.

static int push_samples	(	ATempoContext *	atempo,
		AVFilterLink *	outlink,
		int	n_out
	)

static

Definition at line 1045 of file af_atempo.c.

Referenced by filter_frame(), and request_frame().

static int filter_frame	(	AVFilterLink *	inlink,
		AVFrame *	src_buffer
	)

static

Definition at line 1071 of file af_atempo.c.

static int request_frame ( AVFilterLink * outlink )

static

Definition at line 1112 of file af_atempo.c.

static int process_command	(	AVFilterContext *	ctx,
		const char *	cmd,
		const char *	arg,
		char *	res,
		int	res_len,
		int	flags
	)

static

Definition at line 1156 of file af_atempo.c.

Variable Documentation

const AVOption atempo_options[]

static

Initial value:

= {
    { "tempo", "set tempo scale factor",
      OFFSET(tempo), AV_OPT_TYPE_DOUBLE, { .dbl = 1.0 }, 0.5, 2.0,
      AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_FILTERING_PARAM },
    { NULL }
}

Definition at line 154 of file af_atempo.c.

const AVFilterPad atempo_inputs[]

static

Initial value:

= {
    {
        .name         = "default",
        .type         = AVMEDIA_TYPE_AUDIO,
        .filter_frame = filter_frame,
        .config_props = config_props,
    },
    { NULL }
}

Definition at line 1166 of file af_atempo.c.

const AVFilterPad atempo_outputs[]

static

Initial value:

= {
    {
        .name          = "default",
        .request_frame = request_frame,
        .type          = AVMEDIA_TYPE_AUDIO,
    },
    { NULL }
}

Definition at line 1176 of file af_atempo.c.

AVFilter avfilter_af_atempo

Initial value:

= {
    .name            = "atempo",
    .description     = NULL_IF_CONFIG_SMALL("Adjust audio tempo."),
    .init            = init,
    .uninit          = uninit,
    .query_formats   = query_formats,
    .process_command = process_command,
    .priv_size       = sizeof(ATempoContext),
    .priv_class      = &atempo_class,
    .inputs          = atempo_inputs,
    .outputs         = atempo_outputs,
}

Definition at line 1185 of file af_atempo.c.

Data Structures

Macros

Enumerations

Functions

Variables

Detailed Description

Macro Definition Documentation

Enumeration Type Documentation

Function Documentation

Variable Documentation