search

google / visqol

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

Input sample rate 8kHz #104

Open ray2510 opened 10 months ago

ray2510 commented 10 months ago

Hi, I noticed there are two modes: Audio for 48kHz and Speech for 16kHz. Would the score be accurate if both the reference and degraded samples were at an 8kHz sample rate in speech mode?

I received this message when I attempted to use an 8kHz sample rate for both the reference and degraded samples. [equivalent_rectangular_bandwidth.cc : 33] RAW: EquivalentRectangularBandwidth::MakeFilters high_freq >= (sample_rate / 2), for sample_rate=8000 high_freq=8000.000000. Falling back to (sample_rate / 2)

I'm unsure about the meaning of 'Use a wideband model (sensitive up to 8kHz) with voice activity detection.' Thanks!

DingWeiPeng commented 9 months ago

Hello, could you please share the binary package of visqol? Building it on linux and window both happens errors. I am a little of crazy

GitYesm commented 8 months ago

Hi, I noticed there are two modes: Audio for 48kHz and Speech for 16kHz. Would the score be accurate if both the reference and degraded samples were at an 8kHz sample rate in speech mode?

I received this message when I attempted to use an 8kHz sample rate for both the reference and degraded samples. [equivalent_rectangular_bandwidth.cc : 33] RAW: EquivalentRectangularBandwidth::MakeFilters high_freq >= (sample_rate / 2), for sample_rate=8000 high_freq=8000.000000. Falling back to (sample_rate / 2)

I'm unsure about the meaning of 'Use a wideband model (sensitive up to 8kHz) with voice activity detection.' Thanks!

I don't think the model is accurate in testing narrow bands