avfilter: add axcorrelate filter

Changelog

@@ -24,6 +24,7 @@ version <next>:
 - AV1 encoding support via librav1e
 - AV1 frame merge bitstream filter
 - AV1 Annex B demuxer
+- axcorrelate filter
 
 
 version 4.2:

doc/filters.texi

@@ -2531,6 +2531,39 @@ ffmpeg -i INPUT -af atrim=end_sample=1000
 
 @end itemize
 
+@section axcorrelate
+Calculate normalized cross-correlation between two input audio streams.
+
+Resulted samples are always between -1 and 1 inclusive.
+If result is 1 it means two input samples are highly correlated in that selected segment.
+Result 0 means they are not correlated at all.
+If result is -1 it means two input samples are out of phase, which means they cancel each
+other.
+
+The filter accepts the following options:
+
+@table @option
+@item size
+Set size of segment over which cross-correlation is calculated.
+Default is 256. Allowed range is from 2 to 131072.
+
+@item algo
+Set algorithm for cross-correlation. Can be @code{slow} or @code{fast}.
+Default is @code{slow}. Fast algorithm assumes mean values over any given segment
+are always zero and thus need much less calculations to make.
+This is generally not true, but is valid for typical audio streams.
+@end table
+
+@subsection Examples
+
+@itemize
+@item
+Calculate correlation between channels in stereo audio stream:
+@example
+ffmpeg -i stereo.wav -af channelsplit,axcorrelate=size=1024:algo=fast correlation.wav
+@end example
+@end itemize
+
 @section bandpass
 
 Apply a two-pole Butterworth band-pass filter with central

libavfilter/Makefile

@@ -88,6 +88,7 @@ OBJS-$(CONFIG_ASTATS_FILTER)                 += af_astats.o
 OBJS-$(CONFIG_ASTREAMSELECT_FILTER)          += f_streamselect.o framesync.o
 OBJS-$(CONFIG_ATEMPO_FILTER)                 += af_atempo.o
 OBJS-$(CONFIG_ATRIM_FILTER)                  += trim.o
+OBJS-$(CONFIG_AXCORRELATE_FILTER)            += af_axcorrelate.o
 OBJS-$(CONFIG_AZMQ_FILTER)                   += f_zmq.o
 OBJS-$(CONFIG_BANDPASS_FILTER)               += af_biquads.o
 OBJS-$(CONFIG_BANDREJECT_FILTER)             += af_biquads.o

libavfilter/af_axcorrelate.c

+/*
+ * Copyright (c) 2019 Paul B Mahol
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/avassert.h"
+#include "libavutil/audio_fifo.h"
+#include "libavutil/channel_layout.h"
+#include "libavutil/common.h"
+#include "libavutil/opt.h"
+
+#include "audio.h"
+#include "avfilter.h"
+#include "formats.h"
+#include "filters.h"
+#include "internal.h"
+
+typedef struct AudioXCorrelateContext {
+    const AVClass *class;
+
+    int size;
+    int algo;
+    int64_t pts;
+
+    AVAudioFifo *fifo[2];
+    AVFrame *cache[2];
+    AVFrame *mean_sum[2];
+    AVFrame *num_sum;
+    AVFrame *den_sum[2];
+    int used;
+
+    int (*xcorrelate)(AVFilterContext *ctx, AVFrame *out);
+} AudioXCorrelateContext;
+
+static int query_formats(AVFilterContext *ctx)
+{
+    AVFilterFormats *formats;
+    AVFilterChannelLayouts *layouts;
+    static const enum AVSampleFormat sample_fmts[] = {
+        AV_SAMPLE_FMT_FLTP,
+        AV_SAMPLE_FMT_NONE
+    };
+    int ret;
+
+    layouts = ff_all_channel_counts();
+    if (!layouts)
+        return AVERROR(ENOMEM);
+    ret = ff_set_common_channel_layouts(ctx, layouts);
+    if (ret < 0)
+        return ret;
+
+    formats = ff_make_format_list(sample_fmts);
+    if (!formats)
+        return AVERROR(ENOMEM);
+    ret = ff_set_common_formats(ctx, formats);
+    if (ret < 0)
+        return ret;
+
+    formats = ff_all_samplerates();
+    if (!formats)
+        return AVERROR(ENOMEM);
+    return ff_set_common_samplerates(ctx, formats);
+}
+
+static float mean_sum(const float *in, int size)
+{
+    float mean_sum = 0.f;
+
+    for (int i = 0; i < size; i++)
+        mean_sum += in[i];
+
+    return mean_sum;
+}
+
+static float square_sum(const float *x, const float *y, int size)
+{
+    float square_sum = 0.f;
+
+    for (int i = 0; i < size; i++)
+        square_sum += x[i] * y[i];
+
+    return square_sum;
+}
+
+static float xcorrelate(const float *x, const float *y, float sumx, float sumy, int size)
+{
+    const float xm = sumx / size, ym = sumy / size;
+    float num = 0.f, den, den0 = 0.f, den1 = 0.f;
+
+    for (int i = 0; i < size; i++) {
+        float xd = x[i] - xm;
+        float yd = y[i] - ym;
+
+        num += xd * yd;
+        den0 += xd * xd;
+        den1 += yd * yd;
+    }
+
+    num /= size;
+    den  = sqrtf((den0 * den1) / (size * size));
+
+    return den <= 1e-6f ? 0.f : num / den;
+}
+
+static int xcorrelate_slow(AVFilterContext *ctx, AVFrame *out)
+{
+    AudioXCorrelateContext *s = ctx->priv;
+    const int size = s->size;
+    int used;
+
+    for (int ch = 0; ch < out->channels; ch++) {
+        const float *x = (const float *)s->cache[0]->extended_data[ch];
+        const float *y = (const float *)s->cache[1]->extended_data[ch];
+        float *sumx = (float *)s->mean_sum[0]->extended_data[ch];
+        float *sumy = (float *)s->mean_sum[1]->extended_data[ch];
+        float *dst = (float *)out->extended_data[ch];
+
+        used = s->used;
+        if (!used) {
+            sumx[0] = mean_sum(x, size);
+            sumy[0] = mean_sum(y, size);
+            used = 1;
+        }
+
+        for (int n = 0; n < out->nb_samples; n++) {
+            dst[n] = xcorrelate(x + n, y + n, sumx[0], sumy[0], size);
+
+            sumx[0] -= x[n];
+            sumx[0] += x[n + size];
+            sumy[0] -= y[n];
+            sumy[0] += y[n + size];
+        }
+    }
+
+    return used;
+}
+
+static int xcorrelate_fast(AVFilterContext *ctx, AVFrame *out)
+{
+    AudioXCorrelateContext *s = ctx->priv;
+    const int size = s->size;
+    int used;
+
+    for (int ch = 0; ch < out->channels; ch++) {
+        const float *x = (const float *)s->cache[0]->extended_data[ch];
+        const float *y = (const float *)s->cache[1]->extended_data[ch];
+        float *num_sum = (float *)s->num_sum->extended_data[ch];
+        float *den_sumx = (float *)s->den_sum[0]->extended_data[ch];
+        float *den_sumy = (float *)s->den_sum[1]->extended_data[ch];
+        float *dst = (float *)out->extended_data[ch];
+
+        used = s->used;
+        if (!used) {
+            num_sum[0]  = square_sum(x, y, size);
+            den_sumx[0] = square_sum(x, x, size);
+            den_sumy[0] = square_sum(y, y, size);
+            used = 1;
+        }
+
+        for (int n = 0; n < out->nb_samples; n++) {
+            float num, den;
+
+            num = num_sum[0] / size;
+            den = sqrtf((den_sumx[0] * den_sumy[0]) / (size * size));
+
+            dst[n] = den <= 1e-6f ? 0.f : num / den;
+
+            num_sum[0]  -= x[n] * y[n];
+            num_sum[0]  += x[n + size] * y[n + size];
+            den_sumx[0] -= x[n] * x[n];
+            den_sumx[0]  = FFMAX(den_sumx[0], 0.f);
+            den_sumx[0] += x[n + size] * x[n + size];
+            den_sumy[0] -= y[n] * y[n];
+            den_sumy[0]  = FFMAX(den_sumy[0], 0.f);
+            den_sumy[0] += y[n + size] * y[n + size];
+        }
+    }
+
+    return used;
+}
+
+static int activate(AVFilterContext *ctx)
+{
+    AudioXCorrelateContext *s = ctx->priv;
+    AVFrame *frame = NULL;
+    int ret, status;
+    int available;
+    int64_t pts;
+
+    FF_FILTER_FORWARD_STATUS_BACK_ALL(ctx->outputs[0], ctx);
+
+    for (int i = 0; i < 2; i++) {
+        ret = ff_inlink_consume_frame(ctx->inputs[i], &frame);
+        if (ret > 0) {
+            if (s->pts == AV_NOPTS_VALUE)
+                s->pts = frame->pts;
+            ret = av_audio_fifo_write(s->fifo[i], (void **)frame->extended_data,
+                                      frame->nb_samples);
+            av_frame_free(&frame);
+            if (ret < 0)
+                return ret;
+        }
+    }
+
+    available = FFMIN(av_audio_fifo_size(s->fifo[0]), av_audio_fifo_size(s->fifo[1]));
+    if (available > s->size) {
+        const int out_samples = available - s->size;
+        AVFrame *out;
+
+        if (!s->cache[0] || s->cache[0]->nb_samples < available) {
+            av_frame_free(&s->cache[0]);
+            s->cache[0] = ff_get_audio_buffer(ctx->outputs[0], available);
+            if (!s->cache[0])
+                return AVERROR(ENOMEM);
+        }
+
+        if (!s->cache[1] || s->cache[1]->nb_samples < available) {
+            av_frame_free(&s->cache[1]);
+            s->cache[1] = ff_get_audio_buffer(ctx->outputs[0], available);
+            if (!s->cache[1])
+                return AVERROR(ENOMEM);
+        }
+
+        ret = av_audio_fifo_peek(s->fifo[0], (void **)s->cache[0]->extended_data, available);
+        if (ret < 0)
+            return ret;;
+
+        ret = av_audio_fifo_peek(s->fifo[1], (void **)s->cache[1]->extended_data, available);
+        if (ret < 0)
+            return ret;;
+
+        out = ff_get_audio_buffer(ctx->outputs[0], out_samples);
+        if (!out)
+            return AVERROR(ENOMEM);
+
+        s->used = s->xcorrelate(ctx, out);
+
+        out->pts = s->pts;
+        s->pts += out_samples;
+
+        av_audio_fifo_drain(s->fifo[0], out_samples);
+        av_audio_fifo_drain(s->fifo[1], out_samples);
+
+        return ff_filter_frame(ctx->outputs[0], out);
+    }
+
+    if (av_audio_fifo_size(s->fifo[0]) > s->size &&
+        av_audio_fifo_size(s->fifo[1]) > s->size) {
+        ff_filter_set_ready(ctx, 10);
+        return 0;
+    }
+
+    for (int i = 0; i < 2; i++) {
+        if (ff_inlink_acknowledge_status(ctx->inputs[i], &status, &pts)) {
+            ff_outlink_set_status(ctx->outputs[0], status, pts);
+            return 0;
+        }
+    }
+
+    if (ff_outlink_frame_wanted(ctx->outputs[0])) {
+        for (int i = 0; i < 2; i++) {
+            if (av_audio_fifo_size(s->fifo[i]) > s->size)
+                continue;
+            ff_inlink_request_frame(ctx->inputs[i]);
+            return 0;
+        }
+    }
+
+    return FFERROR_NOT_READY;
+}
+
+static int config_output(AVFilterLink *outlink)
+{
+    AVFilterContext *ctx = outlink->src;
+    AVFilterLink *inlink = ctx->inputs[0];
+    AudioXCorrelateContext *s = ctx->priv;
+
+    s->pts = AV_NOPTS_VALUE;
+
+    outlink->format = inlink->format;
+    outlink->channels = inlink->channels;
+    s->fifo[0] = av_audio_fifo_alloc(outlink->format, outlink->channels, s->size);
+    s->fifo[1] = av_audio_fifo_alloc(outlink->format, outlink->channels, s->size);
+    if (!s->fifo[0] || !s->fifo[1])
+        return AVERROR(ENOMEM);
+
+    s->mean_sum[0] = ff_get_audio_buffer(outlink, 1);
+    s->mean_sum[1] = ff_get_audio_buffer(outlink, 1);
+    s->num_sum = ff_get_audio_buffer(outlink, 1);
+    s->den_sum[0] = ff_get_audio_buffer(outlink, 1);
+    s->den_sum[1] = ff_get_audio_buffer(outlink, 1);
+    if (!s->mean_sum[0] || !s->mean_sum[1] || !s->num_sum ||
+        !s->den_sum[0] || !s->den_sum[1])
+        return AVERROR(ENOMEM);
+
+    switch (s->algo) {
+    case 0: s->xcorrelate = xcorrelate_slow; break;
+    case 1: s->xcorrelate = xcorrelate_fast; break;
+    }
+
+    return 0;
+}
+
+static av_cold void uninit(AVFilterContext *ctx)
+{
+    AudioXCorrelateContext *s = ctx->priv;
+
+    av_audio_fifo_free(s->fifo[0]);
+    av_audio_fifo_free(s->fifo[1]);
+    av_frame_free(&s->cache[0]);
+    av_frame_free(&s->cache[1]);
+    av_frame_free(&s->mean_sum[0]);
+    av_frame_free(&s->mean_sum[1]);
+    av_frame_free(&s->num_sum);
+    av_frame_free(&s->den_sum[0]);
+    av_frame_free(&s->den_sum[1]);
+}
+
+static const AVFilterPad inputs[] = {
+    {
+        .name = "axcorrelate0",
+        .type = AVMEDIA_TYPE_AUDIO,
+    },
+    {
+        .name = "axcorrelate1",
+        .type = AVMEDIA_TYPE_AUDIO,
+    },
+    { NULL }
+};
+
+static const AVFilterPad outputs[] = {
+    {
+        .name         = "default",
+        .type         = AVMEDIA_TYPE_AUDIO,
+        .config_props = config_output,
+    },
+    { NULL }
+};
+
+#define AF AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
+#define OFFSET(x) offsetof(AudioXCorrelateContext, x)
+
+static const AVOption axcorrelate_options[] = {
+    { "size", "set segment size", OFFSET(size), AV_OPT_TYPE_INT,   {.i64=256}, 2, 131072, AF },
+    { "algo", "set alghorithm",   OFFSET(algo), AV_OPT_TYPE_INT,   {.i64=0},   0,      1, AF, "algo" },
+    { "slow", "slow algorithm",   0,            AV_OPT_TYPE_CONST, {.i64=0},   0,      0, AF, "algo" },
+    { "fast", "fast algorithm",   0,            AV_OPT_TYPE_CONST, {.i64=1},   0,      0, AF, "algo" },
+    { NULL }
+};
+
+AVFILTER_DEFINE_CLASS(axcorrelate);
+
+AVFilter ff_af_axcorrelate = {
+    .name           = "axcorrelate",
+    .description    = NULL_IF_CONFIG_SMALL("Cross-correlate two audio streams."),
+    .priv_size      = sizeof(AudioXCorrelateContext),
+    .priv_class     = &axcorrelate_class,
+    .query_formats  = query_formats,
+    .activate       = activate,
+    .uninit         = uninit,
+    .inputs         = inputs,
+    .outputs        = outputs,
+};

libavfilter/allfilters.c

@@ -81,6 +81,7 @@ extern AVFilter ff_af_astats;
 extern AVFilter ff_af_astreamselect;
 extern AVFilter ff_af_atempo;
 extern AVFilter ff_af_atrim;
+extern AVFilter ff_af_axcorrelate;
 extern AVFilter ff_af_azmq;
 extern AVFilter ff_af_bandpass;
 extern AVFilter ff_af_bandreject;

libavfilter/version.h

@@ -30,7 +30,7 @@
 #include "libavutil/version.h"
 
 #define LIBAVFILTER_VERSION_MAJOR   7
-#define LIBAVFILTER_VERSION_MINOR  66
+#define LIBAVFILTER_VERSION_MINOR  67
 #define LIBAVFILTER_VERSION_MICRO 100