Q1. Can we divide phonemes which are specific to a language into some more fundamental entities same across all languages ?

Q2. Will this help us in achieving language independent acoustic modelling ?

Q3. What are the challenges, If any in above technique ?

Q4. Is there any project using this stuff currently ?

  • $\begingroup$ If I am not mistaking, Q1 is a subfield in linguistics. The point is not if it is doable technically (it is) but if it is valid to generalise across all human languages. My perception is that this question would have better chances of getting useful answers, first at the linguistics SE. $\endgroup$ – A_A Sep 4 '18 at 9:12
  • $\begingroup$ Okay, I've posted the question to linguistics.stackexchange.com/q/28893/22631 $\endgroup$ – humble_wolf Sep 4 '18 at 9:24
  • $\begingroup$ Thanks for getting back about this, you might want to delete this question later on if it doesn't apply to DSP.SE, or, modify it so that it is more about the technical aspect of it rather than the linguistics. On the technical side of things, there are many questions here on the use of MFCC and spectrograms applied to speech processing that you might find useful later on. $\endgroup$ – A_A Sep 4 '18 at 10:35
  • $\begingroup$ I will modify it later $\endgroup$ – humble_wolf Sep 4 '18 at 11:24