Basop 2380 regression energy loss in stereo dtx decoding[split-non-BE][non-BE]
- Related issues: #2380 (closed) , #2381 (closed)
- Requested reviewers: @vaclav
Reason why this change is needed
- Energy drop in noisy high-level region for LTV test case around 02:10. MLD
- Can happen when there is a sharp difference of energy from one frame to another and the Q_exc is not appropriately adapted.
- Reuse Ittiam approach without the intermediate truncation and only for the first computation of the gain as the 2 excitations vector have the same dynamic, there is not need for a second verification of the Q_exc value
Affected operating points
- Encoder should be is BE (as well as EVS)
- All synthesis where GSC is used could be affected.
Edited by vaillancour