avfilter/af_silenceremove: add standard deviation detector

Useful in cases audio samples DC offset is not ~0.0, where other detectors will fail to detect silence.

Paul B Mahol committed 3 years ago

Commit f02964a1 parent aa4acc13 files · +28 −0Browse files

Unified · Split

3 files

doc/filters.texi +2 −0
libavfilter/af_silenceremove.c +9 −0
libavfilter/silenceremove_template.c +17 −0

@@ -6474,6 +6474,8 @@ Maximum of absolute values of samples in moving window. · 6473 unmodified lines
6474	6474	Median of absolute values of samples in moving window.
6475	6475	@item ptp
6476	6476	Absolute of max peak to min peak difference of samples in moving window.
	6477	+@item dev
	6478	+Standard deviation of values of samples in moving window.
6477	6479	@end table
6478	6480	Default value is @code{rms}.
6479	6481

libavfilter/af_silenceremove.c+9−0

@@ -38,6 +38,7 @@ enum SilenceDetect { · 37 unmodified lines
38	38	D_PEAK,
39	39	D_MEDIAN,
40	40	D_PTP,
	41	+ D_DEV,
41	42	D_NB
42	43	};
43	44
@@ -146,6 +147,7 @@ static const AVOption silenceremove_options[] = { · 102 unmodified lines
146	147	{ "peak", "use max absolute values of samples", 0, AV_OPT_TYPE_CONST, {.i64=D_PEAK},0, 0, AF, "detection" },
147	148	{ "median", "use median of absolute values of samples", 0, AV_OPT_TYPE_CONST, {.i64=D_MEDIAN},0, 0, AF, "detection" },
148	149	{ "ptp", "use absolute of max peak to min peak difference", 0, AV_OPT_TYPE_CONST, {.i64=D_PTP}, 0, 0, AF, "detection" },
	150	+ { "dev", "use standard deviation from values of samples", 0, AV_OPT_TYPE_CONST, {.i64=D_DEV}, 0, 0, AF, "detection" },
149	151	{ "window", "set duration of window for silence detection", OFFSET(window_duration_opt), AV_OPT_TYPE_DURATION, {.i64=20000}, 0, 100000000, AF },
150	152	{ "timestamp", "set how every output frame timestamp is processed", OFFSET(timestamp_mode), AV_OPT_TYPE_INT, {.i64=TS_WRITE}, 0, TS_NB-1, AF, "timestamp" },
151	153	{ "write", "full timestamps rewrite, keep only the start time", 0, AV_OPT_TYPE_CONST, {.i64=TS_WRITE}, 0, 0, AF, "timestamp" },
@@ -230,6 +232,9 @@ static int config_output(AVFilterLink *outlink) · 78 unmodified lines
230	232	case D_RMS:
231	233	s->cache_size = 1;
232	234	break;
	235	+ case D_DEV:
	236	+ s->cache_size = 2;
	237	+ break;
233	238	case D_MEDIAN:
234	239	case D_PEAK:
235	240	case D_PTP:
@@ -263,6 +268,10 @@ static int config_output(AVFilterLink *outlink) · 27 unmodified lines
263	268	s->compute_flt = compute_avg_flt;
264	269	s->compute_dbl = compute_avg_dbl;
265	270	break;
	271	+ case D_DEV:
	272	+ s->compute_flt = compute_dev_flt;
	273	+ s->compute_dbl = compute_dev_dbl;
	274	+ break;
266	275	case D_PTP:
267	276	s->compute_flt = compute_ptp_flt;
268	277	s->compute_dbl = compute_ptp_dbl;

libavfilter/silenceremove_template.c+17−0

@@ -307,6 +307,23 @@ static ftype fn(compute_rms)(ftype *cache, ftype sample, ftype wsample, · 306 unmodified lines
307	307	return SQRT(r / window_size);
308	308	}
309	309
	310	+static ftype fn(compute_dev)(ftype *ss, ftype x, ftype px,
	311	+ int n, int unused, int unused2)
	312	+{
	313	+ ftype r;
	314	+
	315	+ ss[0] += x;
	316	+ ss[0] -= px;
	317	+
	318	+ ss[1] += x * x;
	319	+ ss[1] -= px * px;
	320	+ ss[1] = FMAX(ss[1], ZERO);
	321	+
	322	+ r = FMAX(ss[1] - ss[0] * ss[0] / n, ZERO) / n;
	323	+
	324	+ return SQRT(r);
	325	+}
	326	+
310	327	static void fn(filter_start)(AVFilterContext *ctx,
311	328	const ftype src, ftype dst,
312	329	int *nb_out_samples,

avfilter/af_silenceremove: add standard deviation detector

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code