Why gaussian assumption in GMM-HMM ASR?

Question

I am reading a book entitled Speech and Language Processing by Daniel Jurafsky and James.

For Acoustic vector, I would like to know why Gaussian assumption is made?

I searched over the internet and could not find the reason.

Answer 1

One reason Gaussian distributions are assumed is that it simplifies modeling since a Gaussian distribution only has 2 parameters to estimate (i.e., mu and sigma) and the properties of the distribution are well understood.

Gaussian distributions are often a reasonable assumption for acoustic data since acoustic data can modeled as a sum of independent random variables. According to the central limit theorem, the normalized sum of independent random variables tends toward a Gaussian distribution.