Noise estimation using high frequency regions for speech enhancement in low SNR environments

The paper presents a noise estimation method for speech enhancement, which provides a good result in low SNR environments. For low SNR speech, the amount of the noise spectrum components is more than that of the speech spectrum components. By using this property, high frequency components of a noisy speech spectrum are averaged and the noise level is estimated. The noise level estimation is performed in the analysis segment, thus this method can adapt to non-stationary noise conditions. The effectiveness of the proposed noise estimation method is confirmed by experiments.