Degrading a signal

[Alex S]

From the videos, I understand that when we implement unit selection using diphones, then we can use techniques such as TD-PSOLA to manipulate F0 and duration. However, excessive manipulation can introduce perceptible artefacts. You mention in a video that it degrades the signal, can you explain what ‘degrading’ means in this context?
Is it simply inferring a lower quality signal? If so, what does a low quality explicitly mean in the context of speech synthesis?

[Simon King]

The terms “degrading” is somewhat informal, but what I mean is that the low-level signal quality is made worse.

This can be contrasted to other types of degradations – for example, that we might get from unit selection: perceptible joins, incorrect co-articulation or bad prosody.

[Author]

Posts