[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [AUDITORY] Why is it that joint speech-enhancement with ASR is not a popular research topic?



Hi Phil, 

Thanks for your insightful response and pointing me to your duplication on this topic from 2003. 
I am particularly intrigued by your comment,    

I am particularly intrigued with your comment:
" It would be wrong to start with clean speech, add noise, use that as input and clean speech + text as training targets, because in real life speech& other sound sources don't combine like that. "

There are many recent publication on speech enhancement  that are using a simple additive noise model, and sometimes RIR simulator, and they are publishing impressive results. Is there a need to incorporate any thing beyond RIR to generalize the training dataset to create a solution that would work properly in the real world?    

Regards,

Samer    



On Mon, Jun 25, 2018 at 9:13 PM Phil Green <p.green@xxxxxxxxxxxxxxx> wrote:



On 25/06/2018 17:00, Samer Hijazi wrote:
Thanks Laszlo and Phil,
I am not speaking about doing ASR in two steps, i am speaking about doing the ASR and speech enhancement jointly in multi-objective learning process.
Are, you mean multitask learning. That didn't come over at all in your first mail.
There are many papers showing if you used related objective resumes to train your network, you will get better results on both objectives than what you would get if you train for each one separately.
An early paper on this, probably the first application to ASR, was

Parveen & Green, Multitask Learning in Connectionist Robust ASR using Recurrent Neural Networks, Eurospeech 2003.

And it seams obvious that if we used speech contents (i.e. text) and perfect speech waveform as two independent but correlated targets, we will end up with a better text recognition and better speech enhancement; am i missing something?    

It would be wrong to start with clean speech, add noise, use that as input and clean speech + text as training targets, because in real life speech & other sound sources don't combine like that. That's why the spectacular results in the Parveen/Green paper are misleading..

HTH
-- 
*** note email is now p.green@xxxxxxxxxx ***
Professor Phil Green
SPandH
Dept of Computer Science
University of Sheffield
*** note email is now p.green@xxxxxxxxxx ***