Does mixing of speech signals comply with central limit theorem?

In this short communication compliance of central limit theorem (CLT) by the convolutive mixing of two speech signals has been investigated in the frequency domain keeping in view the blind separation of convolutive mixture of speech by the independent component analysis (ICA)-based algorithms exploiting properties of CLT. The convolutive mixing of speech is equivalent to superposition of spectral components of each signal. When two non-Gaussian signals are added, the resulting mixed signal gains Gaussianity over the addend signals under the light of CLT. This paper looks for the similar effect of CLT in the mixing process of speech signals with adequate doubts because speech signals from different speakers differ in spectral contents and are spectrally sparse. Interestingly, the results based on the measure of spectral kurtosis show that mixing process of speech signal does not gain Gaussianity in every frequency bin.