[FFmpeg-devel] [PATCH 1/3] avfilter/af_apsyclip: fix FFT bin indexing

[Jason Jang]

With a complex FFT instead of real FFT, the negative frequencies
are not dropped from the spectrum output, so they need to be scaled
when the positive frequencies are scaled. The location of the top
bin is also different.

Signed-off-by: Jason Jang <jcj83429 at gmail.com>
---
 libavfilter/af_apsyclip.c | 19 +++++++++----------
 1 file changed, 9 insertions(+), 10 deletions(-)

diff --git a/libavfilter/af_apsyclip.c b/libavfilter/af_apsyclip.c
index dc3a8e9..0bc469c 100644
--- a/libavfilter/af_apsyclip.c
+++ b/libavfilter/af_apsyclip.c
@@ -292,10 +292,9 @@ static void calculate_mask_curve(AudioPsyClipContext *s,
         if (i == 0) {
             magnitude = FFABS(spectrum[0]);
         } else if (i == s->fft_size / 2) {
-            magnitude = FFABS(spectrum[1]);
+            magnitude = FFABS(spectrum[s->fft_size]);
         } else {
-            // although the negative frequencies are omitted because they are redundant,
-            // the magnitude of the positive frequencies are not doubled.
+            // Because the input signal is real, the + and - frequencies are redundant.
             // Multiply the magnitude by 2 to simulate adding up the + and - frequencies.
             magnitude = hypotf(spectrum[2 * i], spectrum[2 * i + 1]) * 2;
         }
@@ -315,10 +314,9 @@ static void calculate_mask_curve(AudioPsyClipContext *s,
     for (int i = s->num_psy_bins; i < s->fft_size / 2 + 1; i++) {
         float magnitude;
         if (i == s->fft_size / 2) {
-            magnitude = FFABS(spectrum[1]);
+            magnitude = FFABS(spectrum[s->fft_size]);
         } else {
-            // although the negative frequencies are omitted because they are redundant,
-            // the magnitude of the positive frequencies are not doubled.
+            // Because the input signal is real, the + and - frequencies are redundant.
             // Multiply the magnitude by 2 to simulate adding up the + and - frequencies.
             magnitude = hypotf(spectrum[2 * i], spectrum[2 * i + 1]) * 2;
         }
@@ -360,19 +358,20 @@ static void limit_clip_spectrum(AudioPsyClipContext *s,
     for (int i = 1; i < s->fft_size / 2; i++) {
         float real = clip_spectrum[i * 2];
         float imag = clip_spectrum[i * 2 + 1];
-        // although the negative frequencies are omitted because they are redundant,
-        // the magnitude of the positive frequencies are not doubled.
+        // Because the input signal is real, the + and - frequencies are redundant.
         // Multiply the magnitude by 2 to simulate adding up the + and - frequencies.
         relative_distortion_level = hypotf(real, imag) * 2 / mask_curve[i];
         if (relative_distortion_level > 1.0) {
             clip_spectrum[i * 2] /= relative_distortion_level;
             clip_spectrum[i * 2 + 1] /= relative_distortion_level;
+            clip_spectrum[s->fft_size * 2 - i * 2] /= relative_distortion_level;
+            clip_spectrum[s->fft_size * 2 - i * 2 + 1] /= relative_distortion_level;
         }
     }
     // bin N/2
-    relative_distortion_level = FFABS(clip_spectrum[1]) / mask_curve[s->fft_size / 2];
+    relative_distortion_level = FFABS(clip_spectrum[s->fft_size]) / mask_curve[s->fft_size / 2];
     if (relative_distortion_level > 1.f)
-        clip_spectrum[1] /= relative_distortion_level;
+        clip_spectrum[s->fft_size] /= relative_distortion_level;
 }
 
 static void r2c(float *buffer, int size)
-- 
2.35.0