如何使用FFmpeg检测视频文件中的静音音频通道/轨道

我需要能够扫描视频文件，在事先不知道文件音频映射的情况下，报告哪些音频轨道/通道是静音的。我在一个包含16个音频通道的文件上尝试了这个FFmpeg（v4.1）命令：

$ ffmpeg -i foo5.mxf -map 0:a -af astats -f null -
...
Stream mapping:
  Stream #0:1 -> #0:0 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:2 -> #0:1 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:3 -> #0:2 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:4 -> #0:3 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:5 -> #0:4 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:6 -> #0:5 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:7 -> #0:6 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:8 -> #0:7 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:9 -> #0:8 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:10 -> #0:9 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:11 -> #0:10 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:12 -> #0:11 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:13 -> #0:12 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:14 -> #0:13 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:15 -> #0:14 (pcm_s24be (native) -> pcm_s16le (native))
  Stream #0:16 -> #0:15 (pcm_s24be (native) -> pcm_s16le (native))
...
[Parsed_astats_0 @ 0x9123c0] Channel: 1
[Parsed_astats_0 @ 0x9123c0] DC offset: 0.000070
[Parsed_astats_0 @ 0x9123c0] Min level: -170426368.000000
[Parsed_astats_0 @ 0x9123c0] Max level: 172195840.000000
[Parsed_astats_0 @ 0x9123c0] Min difference: 0.000000
[Parsed_astats_0 @ 0x9123c0] Max difference: 58281984.000000
[Parsed_astats_0 @ 0x9123c0] Mean difference: 3089081.855667
[Parsed_astats_0 @ 0x9123c0] RMS difference: 4825348.720329
[Parsed_astats_0 @ 0x9123c0] Peak level dB: -21.918144
[Parsed_astats_0 @ 0x9123c0] RMS level dB: -34.035425
[Parsed_astats_0 @ 0x9123c0] RMS peak dB: -27.320939
[Parsed_astats_0 @ 0x9123c0] RMS trough dB: -48.414915
[Parsed_astats_0 @ 0x9123c0] Crest factor: 4.035190
[Parsed_astats_0 @ 0x9123c0] Flat factor: 0.000000
[Parsed_astats_0 @ 0x9123c0] Peak count: 2
[Parsed_astats_0 @ 0x9123c0] Bit depth: 20/20
[Parsed_astats_0 @ 0x9123c0] Dynamic range: 98.493854
[Parsed_astats_0 @ 0x9123c0] Zero crossings: 10216
[Parsed_astats_0 @ 0x9123c0] Zero crossings rate: 0.042383
[Parsed_astats_0 @ 0x9123c0] Overall
[Parsed_astats_0 @ 0x9123c0] DC offset: 0.000070
[Parsed_astats_0 @ 0x9123c0] Min level: -170426368.000000
[Parsed_astats_0 @ 0x9123c0] Max level: 172195840.000000
[Parsed_astats_0 @ 0x9123c0] Min difference: 0.000000
[Parsed_astats_0 @ 0x9123c0] Max difference: 58281984.000000
[Parsed_astats_0 @ 0x9123c0] Mean difference: 3089081.855667
[Parsed_astats_0 @ 0x9123c0] RMS difference: 4825348.720329
[Parsed_astats_0 @ 0x9123c0] Peak level dB: -21.918144
[Parsed_astats_0 @ 0x9123c0] RMS level dB: -34.035425
[Parsed_astats_0 @ 0x9123c0] RMS peak dB: -27.320939
[Parsed_astats_0 @ 0x9123c0] RMS trough dB: -48.414915
[Parsed_astats_0 @ 0x9123c0] Flat factor: 0.000000
[Parsed_astats_0 @ 0x9123c0] Peak count: 2.000000
[Parsed_astats_0 @ 0x9123c0] Bit depth: 20/20
[Parsed_astats_0 @ 0x9123c0] Number of samples: 241040
[Parsed_astats_0 @ 0x4def40] Channel: 1
[Parsed_astats_0 @ 0x4def40] DC offset: 0.000070
[Parsed_astats_0 @ 0x4def40] Min level: -170426368.000000
[Parsed_astats_0 @ 0x4def40] Max level: 172199936.000000
[Parsed_astats_0 @ 0x4def40] Min difference: 0.000000
[Parsed_astats_0 @ 0x4def40] Max difference: 58277888.000000
[Parsed_astats_0 @ 0x4def40] Mean difference: 3089075.891088
[Parsed_astats_0 @ 0x4def40] RMS difference: 4825346.346176
[Parsed_astats_0 @ 0x4def40] Peak level dB: -21.917938
[Parsed_astats_0 @ 0x4def40] RMS level dB: -34.035425
[Parsed_astats_0 @ 0x4def40] RMS peak dB: -27.320942
[Parsed_astats_0 @ 0x4def40] RMS trough dB: -48.414908
[Parsed_astats_0 @ 0x4def40] Crest factor: 4.035287
[Parsed_astats_0 @ 0x4def40] Flat factor: 0.000000
[Parsed_astats_0 @ 0x4def40] Peak count: 2
[Parsed_astats_0 @ 0x4def40] Bit depth: 20/20
[Parsed_astats_0 @ 0x4def40] Dynamic range: 98.494061
[Parsed_astats_0 @ 0x4def40] Zero crossings: 10220
[Parsed_astats_0 @ 0x4def40] Zero crossings rate: 0.042400
[Parsed_astats_0 @ 0x4def40] Overall
[Parsed_astats_0 @ 0x4def40] DC offset: 0.000070
[Parsed_astats_0 @ 0x4def40] Min level: -170426368.000000
[Parsed_astats_0 @ 0x4def40] Max level: 172199936.000000
[Parsed_astats_0 @ 0x4def40] Min difference: 0.000000
[Parsed_astats_0 @ 0x4def40] Max difference: 58277888.000000
[Parsed_astats_0 @ 0x4def40] Mean difference: 3089075.891088
[Parsed_astats_0 @ 0x4def40] RMS difference: 4825346.346176
[Parsed_astats_0 @ 0x4def40] Peak level dB: -21.917938
[Parsed_astats_0 @ 0x4def40] RMS level dB: -34.035425
[Parsed_astats_0 @ 0x4def40] RMS peak dB: -27.320942
[Parsed_astats_0 @ 0x4def40] RMS trough dB: -48.414908
[Parsed_astats_0 @ 0x4def40] Flat factor: 0.000000
[Parsed_astats_0 @ 0x4def40] Peak count: 2.000000
[Parsed_astats_0 @ 0x4def40] Bit depth: 20/20
[Parsed_astats_0 @ 0x4def40] Number of samples: 241040
...

通过管道传给grep（再加一点过滤），我可以汇总这些信息并检查音频通道的峰值电平：

$ ffmpeg -i foo5.mxf -map 0:a -af astats -f null - 2>&1 | grep -E "(Channel:|Peak level|Overall)"
[Parsed_astats_0 @ 0x9123c0] Channel: 1
[Parsed_astats_0 @ 0x9123c0] Peak level dB: -21.918144
[Parsed_astats_0 @ 0x9123c0] Overall
[Parsed_astats_0 @ 0x9123c0] Peak level dB: -21.918144
[Parsed_astats_0 @ 0x4def40] Channel: 1
[Parsed_astats_0 @ 0x4def40] Peak level dB: -21.917938
[Parsed_astats_0 @ 0x4def40] Overall
[Parsed_astats_0 @ 0x4def40] Peak level dB: -21.917938
[Parsed_astats_0 @ 0xb92180] Channel: 1
[Parsed_astats_0 @ 0xb92180] Peak level dB: -6153.053111
[Parsed_astats_0 @ 0xb92180] Overall
[Parsed_astats_0 @ 0xb92180] Peak level dB: -6153.053111
...

这样如果我过滤掉峰值电平小于-120dB的结果，基本能得到我需要的信息。但问题是它把所有通道都标识为“Channel 1”——可能是因为每个音频流本身只有一个通道。但我在原始输出里找不到任何能把每个“音频统计区段”（也就是包含从“Channel:”到“Zero crossings rate”的部分）和对应的流（比如0:0、0:1等）关联起来的信息。我能不能只根据输出的顺序来可靠地推断这种对应关系？

或者，我是不是应该使用不同的滤镜（而不是astats）或者参数设置来获取这些信息？

备注：内容来源于stack exchange，提问作者kfank