aspectralstats returns wrong centroid value

[penguin101]

Summary of the bug:

I was running aspectralstats on a file with multiple audio channels. At the beginning of the file, there is only silence so the audio channels returned lavfi.aspectralstats.1.centroid=1, apart two of them that returned lavfi.aspectralstats.1.centroid=11924.6.

I removed options and channels one by one till I arrived to the minimum to reproduce.

How to reproduce:

ffmpeg -v 9 -loglevel 99 -t 1 -i /Users/penguin/Downloads/wav_only.wav -filter_complex "aspectralstats=win_size=65536:overlap=0:win_func=welch,ametadata=print:file=-" -f null -

ffmpeg version 6.1 Copyright (c) 2000-2023 the FFmpeg developers

built with Apple clang version 15.0.0 (clang-1500.0.40.1)
configuration: --prefix=/opt/homebrew/Cellar/ffmpeg/6.1-with-options_3 --enable-shared --cc=clang --host-cflags= --host-ldflags='-Wl,-ld_classic' --enable-gpl --enable-libaom --enable-libdav1d --enable-libmp3lame --enable-libopus --enable-libsnappy --enable-libtheora --enable-libvorbis --enable-libvpx --enable-libx264 --enable-libx265 --enable-libfontconfig --enable-libfreetype --enable-frei0r --enable-libass --enable-demuxer=dash --enable-opencl --enable-audiotoolbox --enable-videotoolbox --enable-neon --disable-htmlpages --enable-libfdk-aac --enable-nonfree
libavutil 58. 29.100 / 58. 29.100
libavcodec 60. 31.102 / 60. 31.102
libavformat 60. 16.100 / 60. 16.100
libavdevice 60. 3.100 / 60. 3.100
libavfilter 9. 12.100 / 9. 12.100
libswscale 7. 5.100 / 7. 5.100
libswresample 4. 12.100 / 4. 12.100
libpostproc 57. 3.100 / 57. 3.100

Splitting the commandline.
Reading option '-v' ... matched as option 'v' (set logging level) with argument '9'.
Reading option '-loglevel' ... matched as option 'loglevel' (set logging level) with argument '99'.
Reading option '-t' ... matched as option 't' (record or transcode "duration" seconds of audio/video) with argument '1'.
Reading option '-i' ... matched as output url with argument '/Users/penguin/Downloads/wav_only.wav'.
Reading option '-filter_complex' ... matched as option 'filter_complex' (create a complex filtergraph) with argument 'aspectralstats=win_size=65536:overlap=0:win_func=welch,ametadata=print:file=-'.
Reading option '-f' ... matched as option 'f' (force format) with argument 'null'.
Reading option '-' ... matched as output url.
Finished splitting the commandline.
Parsing a group of options: global .
Applying option v (set logging level) with argument 9.
Applying option filter_complex (create a complex filtergraph) with argument aspectralstats=win_size=65536:overlap=0:win_func=welch,ametadata=print:file=-.
[AVFilterGraph @ 0x14361a130] Setting 'win_size' to value '65536'
[AVFilterGraph @ 0x14361a130] Setting 'overlap' to value '0'
[AVFilterGraph @ 0x14361a130] Setting 'win_func' to value 'welch'
[AVFilterGraph @ 0x14361a130] Setting 'mode' to value 'print'
[AVFilterGraph @ 0x14361a130] Setting 'file' to value '-'
[pipe @ 0x14361a650] Setting default whitelist 'crypto,data'
[AVIOContext @ 0x14361a7d0] Statistics: 0 bytes written, 0 seeks, 0 writeouts
Successfully parsed a group of options.
Parsing a group of options: input url /Users/penguin/Downloads/wav_only.wav.
Applying option t (record or transcode "duration" seconds of audio/video) with argument 1.
Successfully parsed a group of options.
Opening an input file: /Users/penguin/Downloads/wav_only.wav.
[AVFormatContext @ 0x14361a270] Opening '/Users/penguin/Downloads/wav_only.wav' for reading
[file @ 0x14361a9b0] Setting default whitelist 'file,crypto,data'
Probing wav score:99 size:2048
[wav @ 0x14361a270] Format wav probed with size=2048 and score=99
[wav @ 0x14361a270] Before avformat_find_stream_info() pos: 98 bytes read:65634 seeks:1 nb_streams:1
[wav @ 0x14361a270] probing stream 0 pp:32
[wav @ 0x14361a270] probing stream 0 pp:31
[wav @ 0x14361a270] probing stream 0 pp:30
[wav @ 0x14361a270] probing stream 0 pp:29
[wav @ 0x14361a270] probing stream 0 pp:28
[wav @ 0x14361a270] probing stream 0 pp:27
[wav @ 0x14361a270] probing stream 0 pp:26
[wav @ 0x14361a270] probing stream 0 pp:25
[wav @ 0x14361a270] probing stream 0 pp:24
[wav @ 0x14361a270] probing stream 0 pp:23
[wav @ 0x14361a270] probing stream 0 pp:22
[wav @ 0x14361a270] probing stream 0 pp:21
[wav @ 0x14361a270] probing stream 0 pp:20
[wav @ 0x14361a270] probing stream 0 pp:19
[wav @ 0x14361a270] probing stream 0 pp:18
[wav @ 0x14361a270] probing stream 0 pp:17
[wav @ 0x14361a270] probing stream 0 pp:16
[wav @ 0x14361a270] probing stream 0 pp:15
[wav @ 0x14361a270] probing stream 0 pp:14
[wav @ 0x14361a270] probing stream 0 pp:13
[wav @ 0x14361a270] probing stream 0 pp:12
[wav @ 0x14361a270] probing stream 0 pp:11
[wav @ 0x14361a270] probing stream 0 pp:10
[wav @ 0x14361a270] probing stream 0 pp:9
[wav @ 0x14361a270] probing stream 0 pp:8
[wav @ 0x14361a270] probing stream 0 pp:7
[wav @ 0x14361a270] probing stream 0 pp:6
[wav @ 0x14361a270] probing stream 0 pp:5
[wav @ 0x14361a270] probing stream 0 pp:4
[wav @ 0x14361a270] probing stream 0 pp:3
[wav @ 0x14361a270] probing stream 0 pp:2
[wav @ 0x14361a270] probing stream 0 pp:1
[wav @ 0x14361a270] probed stream 0
[wav @ 0x14361a270] parser not found for codec pcm_s16le, packets or times may be invalid.
[wav @ 0x14361a270] All info found
[wav @ 0x14361a270] stream 0: start_time: NOPTS duration: 10
[wav @ 0x14361a270] format: start_time: NOPTS duration: 10 (estimate from stream) bitrate=768 kb/s
[wav @ 0x14361a270] After avformat_find_stream_info() pos: 204898 bytes read:295010 seeks:1 frames:50
[aist#0:0/pcm_s16le @ 0x14361d560] Guessed Channel Layout: mono
Input #0, wav, from '/Users/penguin/Downloads/wav_only.wav':

Metadata:

encoder : Lavf60.16.100
timecode : 00:00:00:00

Duration: 00:00:10.00, bitrate: 768 kb/s
Stream #0:0, 50, 1/48000: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 48000 Hz, 1 channels, s16, 768 kb/s

Successfully opened the file.
Parsing a group of options: output url -.
Applying option f (force format) with argument null.
Successfully parsed a group of options.
Opening an output file: -.
[out#0/null @ 0x144004090] Creating output stream from unlabeled output of complex filtergraph 0. This overrides automatic audio mapping.
[aost#0:0/pcm_s16le @ 0x144004ca0] Created audio stream from complex filtergraph 0:[ametadata:default]
[aost#0:0/pcm_s16le @ 0x144004ca0]
[out#0/null @ 0x144004090] No explicit maps, mapping streams automatically...
Successfully opened the file.
Stream mapping:

Stream #0:0 (pcm_s16le) -> aspectralstats:default
ametadata:default -> Stream #0:0 (pcm_s16le)

Press [q] to stop, ? for help
[AVFilterGraph @ 0x143704e60] Setting 'win_size' to value '65536'
[AVFilterGraph @ 0x143704e60] Setting 'overlap' to value '0'
[AVFilterGraph @ 0x143704e60] Setting 'win_func' to value 'welch'
[AVFilterGraph @ 0x143704e60] Setting 'mode' to value 'print'
[AVFilterGraph @ 0x143704e60] Setting 'file' to value '-'
detected 10 logical cores
[pipe @ 0x143705210] Setting default whitelist 'crypto,data'
[graph_0_in_0_0 @ 0x143705a80] Setting 'time_base' to value '1/48000'
[graph_0_in_0_0 @ 0x143705a80] Setting 'sample_rate' to value '48000'
[graph_0_in_0_0 @ 0x143705a80] Setting 'sample_fmt' to value 's16'
[graph_0_in_0_0 @ 0x143705a80] Setting 'channel_layout' to value 'mono'
[graph_0_in_0_0 @ 0x143705a80] tb:1/48000 samplefmt:s16 samplerate:48000 chlayout:mono
[format_out_0_0 @ 0x143706100] Setting 'sample_fmts' to value 's16'
[trim for input stream 0:0 @ 0x143705b30] auto-inserting filter 'auto_aresample_0' between the filter 'graph_0_in_0_0' and the filter 'trim for input stream 0:0'
[format_out_0_0 @ 0x143706100] auto-inserting filter 'auto_aresample_1' between the filter 'Parsed_ametadata_1' and the filter 'format_out_0_0'
[AVFilterGraph @ 0x143704e60] query_formats: 6 queried, 6 merged, 11 already done, 0 delayed
[auto_aresample_0 @ 0x143706a10] [SWR @ 0x138008000] Using s16p internally between filters
[auto_aresample_0 @ 0x143706a10] ch:1 chl:mono fmt:s16 r:48000Hz -> ch:1 chl:mono fmt:fltp r:48000Hz
For transform of length 65536, forward, fft_float, flags: [aligned, out_of_place], found 6 matches:

1: fft_sr_float_neon - type: fft_float, len: [64, 131072], factor: 2, flags: [aligned, out_of_place], prio: 224
2: fft_pfa_float_c - type: fft_float, len: [6, ∞], factors[2]: [7, 5, 3, 2, any], flags: [unaligned, inplace, out_of_place], prio: 112
3: fft_float_c - type: fft_float, len: [2, ∞], factor: -1, flags: [unaligned, out_of_place], prio: 0
4: fft_inplace_small_float_c - type: fft_float, len: [2, 65536], factor: -1, flags: [unaligned, inplace, out_of_place], prio: -256
5: fft_inplace_float_c - type: fft_float, len: [2, ∞], factor: -1, flags: [unaligned, inplace, out_of_place], prio: -512
6: fft_naive_float_c - type: fft_float, len: [2, ∞], factor: -1, flags: [unaligned, out_of_place], prio: -131072

Transform tree:

fft_sr_float_neon - type: fft_float, len: 65536, factor: 2, flags: [aligned, out_of_place]

[auto_aresample_1 @ 0x143706d60] [SWR @ 0x138020000] Using fltp internally between filters
[auto_aresample_1 @ 0x143706d60] ch:1 chl:mono fmt:fltp r:48000Hz -> ch:1 chl:mono fmt:s16 r:48000Hz
Output #0, null, to 'pipe:':

Metadata:

timecode : 00:00:00:00
encoder : Lavf60.16.100

Stream #0:0, 0, 1/48000: Audio: pcm_s16le, 48000 Hz, mono, s16, 768 kb/s

Metadata:

encoder : Lavc60.31.102 pcm_s16le

[out_0_0 @ 0x143705fa0] EOF on sink link out_0_0:default.
No more output streams to write to, finishing.
[aist#0:0/pcm_s16le @ 0x14361d560] Decoder thread received EOF packet
[aist#0:0/pcm_s16le @ 0x14361d560] Decoder returned EOF, finishing
[aist#0:0/pcm_s16le @ 0x14361d560] Terminating decoder thread
[out#0/null @ 0x144004090] All streams finished
[out#0/null @ 0x144004090] Terminating muxer thread
[out#0/null @ 0x144004090] Output file #0 (pipe:):
[out#0/null @ 0x144004090] Output stream #0:0 (audio): 1 frames encoded (48000 samples); 1 packets muxed (96000 bytes);
[out#0/null @ 0x144004090] Total: 1 packets (96000 bytes) muxed
[out#0/null @ 0x144004090] video:0kB audio:94kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: unknown
size=N/A time=00:00:00.00 bitrate=N/A speed= 0x
frame:0 pts:0 pts_time:0
lavfi.aspectralstats.1.mean=2.38036e-08
lavfi.aspectralstats.1.variance=3.11499e-16
lavfi.aspectralstats.1.centroid=11924.6
lavfi.aspectralstats.1.spread=6956.26
lavfi.aspectralstats.1.skewness=0.000764508
lavfi.aspectralstats.1.kurtosis=1.80617
lavfi.aspectralstats.1.entropy=0.00117855
lavfi.aspectralstats.1.flatness=0.99579
lavfi.aspectralstats.1.crest=0
lavfi.aspectralstats.1.flux=5.36412e-06
lavfi.aspectralstats.1.slope=-4.46234e-10
lavfi.aspectralstats.1.decrease=-0.0270039
lavfi.aspectralstats.1.rolloff=20398.7
[AVIOContext @ 0x143705790] Statistics: 557 bytes written, 0 seeks, 1 writeouts
[in#0/wav @ 0x14361a7d0] Terminating demuxer thread
[in#0/wav @ 0x14361a7d0] Input file #0 (/Users/penguin/Downloads/wav_only.wav):
[in#0/wav @ 0x14361a7d0] Input stream #0:0 (audio): 27 packets read (110592 bytes); 25 frames decoded; 0 decode errors (51200 samples);
[in#0/wav @ 0x14361a7d0] Total: 27 packets (110592 bytes) demuxed
[AVIOContext @ 0x14361a520] Statistics: 295010 bytes read, 1 seeks

[penguin101]

Comes from an inaudible noise that can be removed with filter.