search

google / visqol

Perceptual Quality Estimator for speech and audio
Apache License 2.0
683 stars 124 forks source link

Degraded audio sample rate: 0. #56

Closed seba-plpl closed 2 years ago

seba-plpl commented 3 years ago

Hi, I'm taking my first steps with visqol and have following problem: I have two files (good and bad) both have 16kHz sample rate. When i'm trying to compare this files, i have an error saing that degraded sample has 0 rate, as below:

C:\Users\user1\scoop\apps\bazel\4.2.1\visqol_test\visqol\bazel-bin>visqol.exe --reference_file "C:\Users\user1\scoop\apps\bazel\4.2.1\visqol_test\visqol\testdata\clean_speech\mysample\good_16000_Hz.wav" --degraded_file "C:\Users\user1\scoop\apps\bazel\4.2.1\visqol_test\visqol\testdata\clean_speech\mysample\bad_16000_Hz.wav" --verbose --use_speech_mode --similarity_to_quality_model C:\Users\user1\scoop\apps\bazel\4.2.1\visqol_test\visqol\model\libsvm_nu_svr_model.txt [wav_reader.cc : 174] RAW: Error parsing WAV Header - Expected 16bit samples. [misc_audio.cc : 143] RAW: Error reading header for file C:\Users\user1\scoop\apps\bazel\4.2.1\visqol_test\visqol\testdata\clean_speech\mysample\good_16000_Hz.wav. [wav_reader.cc : 174] RAW: Error parsing WAV Header - Expected 16bit samples. [misc_audio.cc : 143] RAW: Error reading header for file C:\Users\user1\scoop\apps\bazel\4.2.1\visqol_test\visqol\testdata\clean_speech\mysample\bad_16000_Hz.wav. [main.cc : 57] RAW: Error executing ViSQOL: INVALID_ARGUMENT: Input audio signals have different sample rates! Reference audio sample rate: 138441597507072. Degraded audio sample rate: 0.

But, after changed sample rate of this files to 48kHz, it works, it can be compared with following result:

"C:\Users\user1\scoop\apps\bazel\4.2.1\visqol_test\visqol\testdata\clean_speech\mysample\good1_16000_Hz.wav" --degraded_file "C:\Users\user1\scoop\apps\bazel\4.2.1\visqol_test\visqol\testdata\clean_speech\mysample\bad1_16000_Hz.wav" --verbose --use_speech_mode --similarity_to_quality_model C:\Users\user1\scoop\apps\bazel\4.2.1\visqol_test\visqol\model\libsvm_nu_svr_model.txt [visqol_manager.cc : 227] RAW: Input audio sample rate is above 16kHz, which may have undesired effects for speech mode. Consider resampling to 16kHz. ViSQOL conformance version: 310 Speech mode

Reference Filepath: C:\Users\user1\scoop\apps\bazel\4.2.1\visqol_test\visqol\testdata\clean_speech\mysample\good1_16000_Hz.wav Degraded Filepath: C:\Users\user1\scoop\apps\bazel\4.2.1\visqol_test\visqol\testdata\clean_speech\mysample\bad1_16000_Hz.wav MOS-LQO: 1.59276

| FVNSIM | Freq Band |

| 0.401463 | 50.000Hz | | 0.490032 | 98.767Hz | | 0.458079 | 156.063Hz | | 0.417346 | 223.380Hz | | 0.361086 | 302.471Hz | | 0.430890 | 395.394Hz | | 0.422619 | 504.570Hz | | 0.387232 | 632.839Hz | | 0.394989 | 783.543Hz | | 0.428883 | 960.604Hz | | 0.433549 | 1168.633Hz | | 0.392224 | 1413.046Hz | | 0.373587 | 1700.205Hz | | 0.323136 | 2037.587Hz | | 0.293950 | 2433.977Hz | | 0.344104 | 2899.694Hz | | 0.340415 | 3446.863Hz | | 0.315738 | 4089.731Hz | | 0.345335 | 4845.034Hz | | 0.382120 | 5732.437Hz | | 0.340726 | 6775.044Hz |

Can you please help me, what i'm doing wrong?

Mentioned files attached mysample.zip

Thans in advance

mchinen commented 2 years ago

Hm, how were these wav files created? Possibly the waveform headers are not being parsed correctly. One workaround would be to output these in audacity or FFMpeg (untransformed) at 16khz again, since those are known to work.

mchinen commented 2 years ago

Closing due to staleness. Feel free to reopen.