Automatic Sound Signals Quality Estimation. Business Benefits. (2/4)
To Audio Codecs Quality Analyzer page
Sevana Voice Quality Measurement Benefits Compared with ITU P.862 PESQ
• PESQ demands test signals to be speech-like because many systems are optimized for speech and respond in unpredictable way to non-speech signals (e.g. tones, noise, ITU-T P.50)
- Sevana solution works with non-speech signals also, one can test different spectrums and “parasite” frequencies in the channel will not affect our system
• PESQ test signal is set by tester and thus vendor estimations may vary from end customer estimations
- Sevana solution does not demand that. Furthermore, our solution can check voice quality directly comparing two audio files, which is a unique feature on the market
• PESQ has the so-called "disturbance processing"
- Sevana solution does not have this, because if signal spectrum is disturbed, that will be reflected in the estimation and will low down its accuracy
We catch audio quality loss when PESQ fails!
The following article describes situation when ITU P.862 standard fails to detect voice quality loss: http://www.microtronix.ca/pesq-disc.html
The article presents a voice file that was equalized such that there is far less low frequency and high frequency energy when compared to the original file thus making it clearly audible that the speech file is degraded when you listen to it, but PESQ reports that the quality between degraded file and the original are the same!
Test results of our software applied to Microtronix files:
• dg001.wav -> 4.7 (4.96) MOS (94% similarity in our units)
• dg002.wav -> 1.29 (0.73) MOS (26% similarity in our units)
This example clearly shows that our software does catch that significant quality loss, which occurs when the file is equalized with far less low frequency and high frequency energy.
|
