Threshold Filtering for Phoneme Pronunciation Signals Based on FrFT

The Fractional Fourier Transform (FrFT) is applied to the denoising of noisy speech. The optimal transform order of FrFT for single phoneme is determined by using weighted variance method. Then the soft-hard threshold compromise denoising algorithm is put forward. This method removes the amplitude of noise from noisy phoneme signals in FrFT-domain. Signals are reconstructed by inverse FrFT to get original speech. The experimental results show that this method can effectively remove noise from signals and get a good auditory effect, and this algorithm is of low computational complexity.