Vocoder: how we can evaluate the performance of a vocoder

[Jiewen Z]

Hi Simon,
I tried the Word vocoder. I used it to parameterize my speech wave and reconstruct speech from parameters. when i hear the constructed voice i found it perfectly reconstructed my voice(because i can not hear any difference between the original wave and the constrcted one). My question is how can i evaluate the performance of a vocoder(method,tool).

Thanks!

[Simon]

If you think that WORLD gives perfect quality when vocoding (which is often called “copy synthesis”) then you either need to listen more closely, or use a better pair of headphones! You should be able to discern some artefacts of vocoding.

To evaluate copy synthesis, a MUSHRA listening test would be most appropriate. The hidden reference would be the original waveform and the lower anchor could be low-pass filtered speech.

[Jiewen Z]

Got it! i think i just did the experiment with my loudspeaker on my laptop,which is hard to detect differences.

After i change to my headphone. i do detect two differences:
1. My voice sometimes become trembling.
2. After the reconstruction, it seems that the background noise from the original wave is amplified and randomized,which make it become more detectable.

[Author]

Posts