Go to the documentation of this file.
33 #define MEASURE_ALL UINT_MAX
34 #define MEASURE_NONE 0
35 #define MEASURE_MEAN (1 << 0)
36 #define MEASURE_VARIANCE (1 << 1)
37 #define MEASURE_CENTROID (1 << 2)
38 #define MEASURE_SPREAD (1 << 3)
39 #define MEASURE_SKEWNESS (1 << 4)
40 #define MEASURE_KURTOSIS (1 << 5)
41 #define MEASURE_ENTROPY (1 << 6)
42 #define MEASURE_FLATNESS (1 << 7)
43 #define MEASURE_CREST (1 << 8)
44 #define MEASURE_FLUX (1 << 9)
45 #define MEASURE_SLOPE (1 << 10)
46 #define MEASURE_DECREASE (1 << 11)
47 #define MEASURE_ROLLOFF (1 << 12)
84 #define OFFSET(x) offsetof(AudioSpectralStatsContext, x)
85 #define A AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
115 float overlap,
scale = 1.f;
120 sizeof(*
s->window_func_lut));
121 if (!
s->window_func_lut)
124 if (
s->overlap == 1.f)
125 s->overlap = overlap;
127 s->hop_size =
s->win_size * (1.f -
s->overlap);
128 if (
s->hop_size <= 0)
131 s->stats =
av_calloc(
s->nb_channels,
sizeof(*
s->stats));
139 s->magnitude =
av_calloc(
s->nb_channels,
sizeof(*
s->magnitude));
143 s->prev_magnitude =
av_calloc(
s->nb_channels,
sizeof(*
s->prev_magnitude));
144 if (!
s->prev_magnitude)
147 s->fft_in =
av_calloc(
s->nb_channels,
sizeof(*
s->fft_in));
151 s->fft_out =
av_calloc(
s->nb_channels,
sizeof(*
s->fft_out));
155 for (
int ch = 0; ch <
s->nb_channels; ch++) {
160 s->fft_in[ch] =
av_calloc(
s->win_size,
sizeof(**
s->fft_in));
164 s->fft_out[ch] =
av_calloc(
s->win_size,
sizeof(**
s->fft_out));
168 s->magnitude[ch] =
av_calloc(
s->win_size,
sizeof(**
s->magnitude));
169 if (!
s->magnitude[ch])
172 s->prev_magnitude[ch] =
av_calloc(
s->win_size,
sizeof(**
s->prev_magnitude));
173 if (!
s->prev_magnitude[ch])
185 const char *fmt,
float val)
192 snprintf(key2,
sizeof(key2),
"lavfi.aspectralstats.%d.%s", chan,
key);
194 snprintf(key2,
sizeof(key2),
"lavfi.aspectralstats.%s",
key);
200 for (
int ch = 0; ch <
s->nb_channels; ch++) {
206 set_meta(metadata, ch + 1,
"variance",
"%g",
stats->variance);
208 set_meta(metadata, ch + 1,
"centroid",
"%g",
stats->centroid);
212 set_meta(metadata, ch + 1,
"skewness",
"%g",
stats->skewness);
214 set_meta(metadata, ch + 1,
"kurtosis",
"%g",
stats->kurtosis);
218 set_meta(metadata, ch + 1,
"flatness",
"%g",
stats->flatness);
226 set_meta(metadata, ch + 1,
"decrease",
"%g",
stats->decrease);
236 for (
int n = 0; n <
size; n++)
251 for (
int n = 0; n <
size; n++)
260 float num = 0.f, den = 0.f;
262 for (
int n = 0; n <
size; n++) {
263 num += spectral[n] * n *
scale;
267 if (den <= FLT_EPSILON)
275 float num = 0.f, den = 0.f;
277 for (
int n = 0; n <
size; n++) {
278 num += spectral[n] *
sqrf(n *
scale - centroid);
282 if (den <= FLT_EPSILON)
284 return sqrtf(num / den);
295 float num = 0.f, den = 0.f;
297 for (
int n = 0; n <
size; n++) {
298 num += spectral[n] *
cbrf(n *
scale - centroid);
303 if (den <= FLT_EPSILON)
311 float num = 0.f, den = 0.f;
313 for (
int n = 0; n <
size; n++) {
319 if (den <= FLT_EPSILON)
326 float num = 0.f, den = 0.f;
328 for (
int n = 0; n <
size; n++) {
329 num += spectral[n] * logf(spectral[n] + FLT_EPSILON);
333 if (den <= FLT_EPSILON)
340 float num = 0.f, den = 0.f;
342 for (
int n = 0; n <
size; n++) {
343 float v = FLT_EPSILON + spectral[n];
351 if (den <= FLT_EPSILON)
360 for (
int n = 0; n <
size; n++) {
366 if (
mean <= FLT_EPSILON)
371 static float spectral_flux(
const float *
const spectral,
const float *
const prev_spectral,
372 int size,
int max_freq)
376 for (
int n = 0; n <
size; n++)
377 sum +=
sqrf(spectral[n] - prev_spectral[n]);
384 const float mean_freq =
size * 0.5f;
385 float mean_spectral = 0.f, num = 0.f, den = 0.f;
387 for (
int n = 0; n <
size; n++)
388 mean_spectral += spectral[n];
389 mean_spectral /=
size;
391 for (
int n = 0; n <
size; n++) {
392 num += ((n - mean_freq) / mean_freq) * (spectral[n] - mean_spectral);
393 den +=
sqrf((n - mean_freq) / mean_freq);
396 if (
fabsf(den) <= FLT_EPSILON)
403 float num = 0.f, den = 0.f;
405 for (
int n = 1; n <
size; n++) {
406 num += (spectral[n] - spectral[0]) / n;
410 if (den <= FLT_EPSILON)
418 float norm = 0.f, sum = 0.f;
421 for (
int n = 0; n <
size; n++)
425 for (
int n = 0; n <
size; n++) {
439 const float *window_func_lut =
s->window_func_lut;
442 const int start = (
channels * jobnr) / nb_jobs;
443 const int end = (
channels * (jobnr+1)) / nb_jobs;
444 const int offset =
s->win_size -
s->hop_size;
446 for (
int ch = start; ch < end; ch++) {
447 float *
window = (
float *)
s->window->extended_data[ch];
451 float *magnitude =
s->magnitude[ch];
452 float *prev_magnitude =
s->prev_magnitude[ch];
453 const float scale = 1.f /
s->win_size;
459 for (
int n = 0; n <
s->win_size; n++) {
460 fft_in[n].re =
window[n] * window_func_lut[n];
464 s->tx_fn(
s->fft[ch], fft_out, fft_in,
sizeof(*fft_in));
466 for (
int n = 0; n <
s->win_size / 2; n++) {
467 fft_out[n].re *=
scale;
468 fft_out[n].im *=
scale;
471 for (
int n = 0; n <
s->win_size / 2; n++)
472 magnitude[n] = hypotf(fft_out[n].re, fft_out[n].im);
501 memcpy(prev_magnitude, magnitude,
s->win_size *
sizeof(
float));
532 metadata = &
out->metadata;
580 for (
int ch = 0; ch <
s->nb_channels; ch++) {
589 if (
s->prev_magnitude)
613 .
name =
"aspectralstats",
616 .priv_class = &aspectralstats_class,
AVFrame * ff_get_audio_buffer(AVFilterLink *link, int nb_samples)
Request an audio samples buffer with a specific set of permissions.
@ AV_SAMPLE_FMT_FLTP
float, planar
static float spectral_mean(const float *const spectral, int size, int max_freq)
Filter the word “frame” indicates either a video frame or a group of audio as stored in an AVFrame structure Format for each input and each output the list of supported formats For video that means pixel format For audio that means channel sample they are references to shared objects When the negotiation mechanism computes the intersection of the formats supported at each end of a all references to both lists are replaced with a reference to the intersection And when a single format is eventually chosen for a link amongst the remaining all references to the list are updated That means that if a filter requires that its input and output have the same format amongst a supported all it has to do is use a reference to the same list of formats query_formats can leave some formats unset and return AVERROR(EAGAIN) to cause the negotiation mechanism toagain later. That can be used by filters with complex requirements to use the format negotiated on one link to set the formats supported on another. Frame references ownership and permissions
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
#define FILTER_SINGLE_SAMPLEFMT(sample_fmt_)
The exact code depends on how similar the blocks are and how related they are to the and needs to apply these operations to the correct inlink or outlink if there are several Macros are available to factor that when no extra processing is inlink
void av_frame_free(AVFrame **frame)
Free the frame and any dynamically allocated objects in it, e.g.
This structure describes decoded (raw) audio or video data.
#define WIN_FUNC_OPTION(win_func_opt_name, win_func_offset, flag, default_window_func)
const char * name
Filter name.
int nb_channels
Number of channels in this layout.
static const AVFilterPad aspectralstats_outputs[]
A link between two filters.
#define FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink)
Forward the status on an output link to an input link.
av_cold int av_tx_init(AVTXContext **ctx, av_tx_fn *tx, enum AVTXType type, int inv, int len, const void *scale, uint64_t flags)
Initialize a transform context with the given configuration (i)MDCTs with an odd length are currently...
static float spectral_entropy(const float *const spectral, int size, int max_freq)
AVComplexFloat ** fft_out
static SDL_Window * window
void * priv
private data for use by the filter
static float spectral_variance(const float *const spectral, int size, int max_freq, float mean)
static double val(void *priv, double ch)
static __device__ float fabsf(float a)
A filter pad used for either input or output.
const AVFilter ff_af_aspectralstats
static float spectral_crest(const float *const spectral, int size, int max_freq)
static int filter_channel(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
void(* av_tx_fn)(AVTXContext *s, void *out, void *in, ptrdiff_t stride)
Function pointer to a function to perform the transform.
static float spectral_flatness(const float *const spectral, int size, int max_freq)
@ AV_TX_FLOAT_FFT
Standard complex to complex FFT with sample data type of AVComplexFloat, AVComplexDouble or AVComplex...
#define FILTER_INPUTS(array)
#define av_realloc_f(p, o, n)
static float spectral_slope(const float *const spectral, int size, int max_freq)
Describe the class of an AVClass context structure.
int ff_inlink_consume_samples(AVFilterLink *link, unsigned min, unsigned max, AVFrame **rframe)
Take samples from the link's FIFO and update the link's stats.
int av_frame_copy_props(AVFrame *dst, const AVFrame *src)
Copy only "metadata" fields from src to dst.
const AVFilterPad ff_audio_default_filterpad[1]
An AVFilterPad array whose only entry has name "default" and is of type AVMEDIA_TYPE_AUDIO.
static __device__ float sqrtf(float a)
static void generate_window_func(float *lut, int N, int win_func, float *overlap)
static const AVOption aspectralstats_options[]
static void stats(AVPacket *const *in, int n_in, unsigned *_max, unsigned *_sum)
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification.
int av_frame_copy(AVFrame *dst, const AVFrame *src)
Copy the frame data from src to dst.
int sample_rate
Sample rate of the audio data.
float fmaxf(float, float)
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
int av_frame_is_writable(AVFrame *frame)
Check if the frame data is writable.
AVFilterContext * src
source filter
The reader does not expect b to be semantically here and if the code is changed by maybe adding a a division or other the signedness will almost certainly be mistaken To avoid this confusion a new type was SUINT is the C unsigned type but it holds a signed int to use the same example SUINT a
static float spectral_decrease(const float *const spectral, int size, int max_freq)
it s the only field you need to keep assuming you have a context There is some magic you don t need to care about around this just let it vf offset
FF_FILTER_FORWARD_WANTED(outlink, inlink)
static int activate(AVFilterContext *ctx)
static float spectral_centroid(const float *const spectral, int size, int max_freq)
av_cold void av_tx_uninit(AVTXContext **ctx)
Frees a context and sets *ctx to NULL, does nothing when *ctx == NULL.
static av_cold void uninit(AVFilterContext *ctx)
static float spectral_rolloff(const float *const spectral, int size, int max_freq)
int nb_samples
number of audio samples (per channel) described by this frame
uint8_t ** extended_data
pointers to the data planes/channels.
static float cbrf(float a)
static float spectral_flux(const float *const spectral, const float *const prev_spectral, int size, int max_freq)
int ff_filter_get_nb_threads(AVFilterContext *ctx)
Get number of threads for current filter instance.
it s the only field you need to keep assuming you have a context There is some magic you don t need to care about around this just let it vf default value
const char * name
Pad name.
int ff_inlink_queued_samples(AVFilterLink *link)
void * av_calloc(size_t nmemb, size_t size)
static int config_output(AVFilterLink *outlink)
static void set_meta(AVDictionary **metadata, int chan, const char *key, const char *fmt, float val)
static float spectral_spread(const float *const spectral, int size, int max_freq, float centroid)
static float mean(const float *input, int size)
static float spectral_skewness(const float *const spectral, int size, int max_freq, float centroid, float spread)
#define AVFILTER_FLAG_SLICE_THREADS
The filter supports multithreading by splitting frames into multiple parts and processing them concur...
AVChannelLayout ch_layout
channel layout of current buffer (see libavutil/channel_layout.h)
static void scale(int *out, const int *in, const int w, const int h, const int shift)
FF_FILTER_FORWARD_STATUS(inlink, outlink)
#define FILTER_OUTPUTS(array)
int av_dict_set(AVDictionary **pm, const char *key, const char *value, int flags)
Set the given entry in *pm, overwriting an existing entry.
static void set_metadata(AudioSpectralStatsContext *s, AVDictionary **metadata)
static float sqrf(float a)
ChannelSpectralStats * stats
AVFILTER_DEFINE_CLASS(aspectralstats)
static av_always_inline int ff_filter_execute(AVFilterContext *ctx, avfilter_action_func *func, void *arg, int *ret, int nb_jobs)
static float spectral_kurtosis(const float *const spectral, int size, int max_freq, float centroid, float spread)
void ff_filter_set_ready(AVFilterContext *filter, unsigned priority)
Mark a filter ready and schedule it for activation.