search

introlab / odas

ODAS: Open embeddeD Audition System
MIT License
780 stars 248 forks source link

Some question about mode of sss. #70

Closed huotuichang1 closed 6 years ago

huotuichang1 commented 6 years ago

Excuse me, I see there is mvdr in the sss_mode but there is not any code.why?Is it not effective?

FrancoisGrondin commented 6 years ago

It is not implemented yet. MVDR uses a lot of computing resources, and was not real time on a raspberry pi. Is this something you'd need?

huotuichang1 commented 6 years ago

you are great~thank you~

taospartan commented 6 years ago

What does MVDR stand for?

Sent from my iPhone

On 9 Jun 2018, at 22:37, François Grondin notifications@github.com wrote:

MVDR

taospartan commented 6 years ago

It’s ok, googled it and found out all about it, from experience is a mvdr beamformer better than a differential for sound source isolation ?

Sent from my iPhone

On 10 Jun 2018, at 14:31, Rich Gunton richgunton@gmail.com wrote:

What does MVDR stand for?

Sent from my iPhone

On 9 Jun 2018, at 22:37, François Grondin notifications@github.com wrote:

MVDR

FrancoisGrondin commented 6 years ago

A MVDR is good to eliminate "equally" all sounds coming from other directions than the desired one. From my limited experience with MVDR, here are the two drawbacks I observed:

1) When you have strong reflections generated by the desired source, these are considered as undesirable interference, and the MVDR tries to attenuate frequency bands that shouldn't be attenuated as they contain the target signal information. 2) You need to compute and invert the correlation matrix at each frequency band, which requires significant computing resources