[FFmpeg-devel,v7] avfilter/avf_aphasemeter: Add out-of-phase and mono detection

Submitted by Romane Lafon on Aug. 19, 2019, 3:37 p.m.

Details

Message ID CAMe7PWRGLLj79RLS-n-53Zsa6fBMjMRfo-tzpG=r+Cbo7RTyHw@mail.gmail.com
State New
Headers show

Commit Message

Romane Lafon Aug. 19, 2019, 3:37 p.m.
This patch extends aphasemeter to detect out of phase or mono sequences in
stereo streams, with its associated documentation.

Patch hide | download patch | download mbox

From dce30e5a978ff5296fa1c0013bbd78fd27ecec2a Mon Sep 17 00:00:00 2001
From: Romane Lafon <romane@nomalab.com>
Date: Fri, 16 Aug 2019 14:14:11 +0200
Subject: [PATCH] avfilter/avf_aphasemeter: Add out-of-phase and mono detection

Signed-off-by: Romane Lafon <romane@nomalab.com>
---
 doc/filters.texi              |  33 ++++++++++
 libavfilter/avf_aphasemeter.c | 121 ++++++++++++++++++++++++++++++++--
 2 files changed, 150 insertions(+), 4 deletions(-)

diff --git a/doc/filters.texi b/doc/filters.texi
index 01262d845e..5acd36a98c 100644
--- a/doc/filters.texi
+++ b/doc/filters.texi
@@ -20822,6 +20822,39 @@  Set color which will be used for drawing median phase. If color is
 Enable video output. Default is enabled.
 @end table
 
+@subsection phasing detection
+
+The filter also detects out of phase and mono sequences in stereo streams.
+It logs the sequence start, end and duration when it lasts longer or as long as the minimum set.
+
+The filter accepts the following options for this detection:
+
+@table @option
+@item phasing
+Enable mono and out of phase detection. Default is disabled.
+
+@item tolerance
+Set phase tolerance for mono detection, in amplitude ratio. Default is @code{0}.
+Allowed range is @code{[0, 1]}.
+
+@item angle
+Set angle threshold for out of phase detection, in degree. Default is @code{170}.
+Allowed range is @code{[0, 180]}.
+
+@item duration
+Set mono or out of phase duration until notification, expressed in seconds. Default is @code{2}.
+@end table
+
+@subsection Examples
+
+@itemize
+@item
+Complete example with @command{ffmpeg} to detect 1 second of mono with 0.001 phase tolerance:
+@example
+ffmpeg -i stereo.wav -af aphasemeter=video=0:phasing=1:duration=1:tolerance=0.001 -f null -
+@end example
+@end itemize
+
 @section avectorscope
 
 Convert input audio to a video output, representing the audio vector
diff --git a/libavfilter/avf_aphasemeter.c b/libavfilter/avf_aphasemeter.c
index f497bc9969..b94cbbade6 100644
--- a/libavfilter/avf_aphasemeter.c
+++ b/libavfilter/avf_aphasemeter.c
@@ -28,26 +28,40 @@ 
 #include "libavutil/intreadwrite.h"
 #include "libavutil/opt.h"
 #include "libavutil/parseutils.h"
+#include "libavutil/timestamp.h"
 #include "avfilter.h"
 #include "formats.h"
 #include "audio.h"
 #include "video.h"
 #include "internal.h"
+#include "float.h"
 
 typedef struct AudioPhaseMeterContext {
     const AVClass *class;
     AVFrame *out;
     int do_video;
+    int do_phasing_detection;
     int w, h;
     AVRational frame_rate;
     int contrast[4];
     uint8_t *mpc_str;
     uint8_t mpc[4];
     int draw_median_phase;
+    int is_mono;
+    int is_out_phase;
+    int start_mono_presence;
+    int start_out_phase_presence;
+    float tolerance;
+    float angle;
+    float phase;
+    float mono_idx[2];
+    float out_phase_idx[2];
+    double duration;
 } AudioPhaseMeterContext;
 
 #define OFFSET(x) offsetof(AudioPhaseMeterContext, x)
 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
+#define get_duration(index) (index[1] - index[0])
 
 static const AVOption aphasemeter_options[] = {
     { "rate", "set video rate", OFFSET(frame_rate), AV_OPT_TYPE_VIDEO_RATE, {.str="25"}, 0, INT_MAX, FLAGS },
@@ -59,6 +73,10 @@  static const AVOption aphasemeter_options[] = {
     { "bc", "set blue contrast",  OFFSET(contrast[2]), AV_OPT_TYPE_INT, {.i64=1}, 0, 255, FLAGS },
     { "mpc", "set median phase color", OFFSET(mpc_str), AV_OPT_TYPE_STRING, {.str = "none"}, 0, 0, FLAGS },
     { "video", "set video output", OFFSET(do_video), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, FLAGS },
+    { "phasing", "set mono and out-of-phase detection output", OFFSET(do_phasing_detection), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, FLAGS },
+    { "tolerance", "set phase tolerance for mono detection", OFFSET(tolerance), AV_OPT_TYPE_FLOAT, {.dbl = 0.}, 0, 1, FLAGS },
+    { "angle", "set angle threshold for out-of-phase detection", OFFSET(angle), AV_OPT_TYPE_FLOAT, {.dbl = 170.}, 90, 180, FLAGS },
+    { "duration", "set minimum mono or out-of-phase duration in seconds", OFFSET(duration), AV_OPT_TYPE_DOUBLE, {.dbl=2.}, 0, 24*60*60, FLAGS },
     { NULL }
 };
 
@@ -140,6 +158,21 @@  static inline int get_x(float phase, int w)
   return (phase + 1.) / 2. * (w - 1);
 }
 
+static inline float get_index(AVFilterLink *inlink, int64_t pts)
+{
+    return atof(av_ts2timestr(pts, &inlink->time_base));
+}
+
+static inline void add_metadata(AVFrame *insamples, const char *key, float value)
+{
+    char buf[128];
+    char str[128];
+
+    snprintf(str, sizeof(str), "%f", value);
+    snprintf(buf, sizeof(buf), "lavfi.aphasemeter.%s", key);
+    av_dict_set(&insamples->metadata, buf, str, 0);
+}
+
 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
 {
     AVFilterContext *ctx = inlink->dst;
@@ -154,6 +187,10 @@  static int filter_frame(AVFilterLink *inlink, AVFrame *in)
     AVFrame *out;
     uint8_t *dst;
     int i;
+    int mono_measurement;
+    int out_phase_measurement;
+    float tolerance = 1.0f - s->tolerance;
+    float angle = cosf(s->angle/180.0f*M_PI);
 
     if (s->do_video && (!s->out || s->out->width  != outlink->w ||
                                    s->out->height != outlink->h)) {
@@ -193,7 +230,7 @@  static int filter_frame(AVFilterLink *inlink, AVFrame *in)
         fphase += phase;
     }
     fphase /= in->nb_samples;
-
+    s->phase = fphase;
     if (s->do_video) {
         if (s->draw_median_phase) {
             dst = out->data[0] + get_x(fphase, s->w) * 4;
@@ -206,10 +243,64 @@  static int filter_frame(AVFilterLink *inlink, AVFrame *in)
 
     metadata = &in->metadata;
     if (metadata) {
-        uint8_t value[128];
+        add_metadata(in, "phase", fphase);
+    }
 
-        snprintf(value, sizeof(value), "%f", fphase);
-        av_dict_set(metadata, "lavfi.aphasemeter.phase", value, 0);
+    if (s->do_phasing_detection) {
+        mono_measurement = (tolerance - fphase) < FLT_EPSILON;
+        out_phase_measurement = (angle - fphase) > FLT_EPSILON;
+        if (!s->is_mono && mono_measurement) {
+            s->is_mono = 1;
+            s->start_mono_presence = 1;
+            s->mono_idx[0] = get_index(inlink, in->pts);
+        }
+        if (s->is_mono && mono_measurement && s->start_mono_presence) {
+            float mono_duration;
+            s->mono_idx[1] = get_index(inlink, in->pts);
+            mono_duration = get_duration(s->mono_idx);
+            if (mono_duration >= s->duration) {
+                add_metadata(in, "mono_start", s->mono_idx[0]);
+                av_log(s, AV_LOG_INFO, "mono_start: %f\n", s->mono_idx[0]);
+                s->start_mono_presence = 0;
+            }
+        }
+        if (s->is_mono && !mono_measurement) {
+            float mono_duration;
+            s->mono_idx[1] = get_index(inlink, in->pts);
+            mono_duration = get_duration(s->mono_idx);
+            if (mono_duration > s->duration) {
+                add_metadata(in, "mono_end", s->mono_idx[1]);
+                add_metadata(in, "mono_duration", mono_duration);
+                av_log(s, AV_LOG_INFO, "mono_end: %f | mono_duration: %f\n", s->mono_idx[1], mono_duration);
+            }
+            s->is_mono = 0;
+        }
+        if (!s->is_out_phase && out_phase_measurement) {
+            s->out_phase_idx[0] = get_index(inlink, in->pts);
+            s->is_out_phase = 1;
+            s->start_out_phase_presence = 1;
+        }
+        if (s->is_out_phase && out_phase_measurement && s->start_out_phase_presence) {
+            float out_phase_duration;
+            s->out_phase_idx[1] = get_index(inlink, in->pts);
+            out_phase_duration = get_duration(s->out_phase_idx);
+            if (out_phase_duration >= s->duration) {
+                add_metadata(in, "out_phase_start", s->out_phase_idx[0]);
+                av_log(s, AV_LOG_INFO, "out_phase_start: %f\n", s->out_phase_idx[0]);
+                s->start_out_phase_presence = 0;
+            }
+        }
+        if (s->is_out_phase && !out_phase_measurement) {
+            float out_phase_duration;
+            s->out_phase_idx[1] = get_index(inlink, in->pts);
+            out_phase_duration = get_duration(s->out_phase_idx);
+            if (out_phase_duration > s->duration) {
+                add_metadata(in, "out_phase_end", s->out_phase_idx[1]);
+                add_metadata(in, "out_phase_duration", out_phase_duration);
+                av_log(s, AV_LOG_INFO, "out_phase_end: %f | out_phase_duration: %f\n", s->out_phase_idx[1], out_phase_duration);
+            }
+            s->is_out_phase = 0;
+        }
     }
 
     if (s->do_video) {
@@ -224,6 +315,28 @@  static av_cold void uninit(AVFilterContext *ctx)
     AudioPhaseMeterContext *s = ctx->priv;
     int i;
 
+    if (s->do_phasing_detection) {
+        float mono_duration;
+        float out_phase_duration;
+        AVFilterLink *inlink = ctx->inputs[0];
+
+        if (s->is_mono) {
+            s->mono_idx[1] = get_index(inlink, inlink->current_pts);
+            mono_duration = get_duration(s->mono_idx);
+            if (mono_duration > s->duration) {
+                av_log(s, AV_LOG_INFO, "mono_end: %f | mono_duration: %f\n", s->mono_idx[1], mono_duration);
+            }
+            s->is_mono = 0;
+        }
+        if (s->is_out_phase) {
+            s->out_phase_idx[1] = get_index(inlink, inlink->current_pts);
+            out_phase_duration = get_duration(s->out_phase_idx);
+            if (out_phase_duration > s->duration) {
+                av_log(s, AV_LOG_INFO, "out_phase_end: %f | out_phase_duration: %f\n", s->out_phase_idx[1], out_phase_duration);
+            }
+            s->is_out_phase = 0;
+        }
+    }
     av_frame_free(&s->out);
     for (i = 0; i < ctx->nb_outputs; i++)
         av_freep(&ctx->output_pads[i].name);
-- 
2.20.1