An Impulse Noise Robust Noise Estimation Algorithm Applied for Low Signal-to-Noise Ratio Digital Communication

This paper presents an impulse noise robust noise estimation algorithm in low signal-to-noise ratio (SNR) environments. This method can be integrated into any speech enhancement algorithm which requires a noise power spectral estimation. In contrast to other methods, this approach does not use a voice activity detector, which restricts the tracking capability of the noise estimator in case of low SNR and varying noise spectrum. The noise estimation is updated continuously by smoothing the noisy speech power spectrum with time and frequency dependent parameters. The optimal parameters are derived from computing signal presence probability by tracking power spectral minima in each frequency bin with a look-ahead factor. The proposed algorithm was evaluated at the presence of impulse noise, yielding results which demonstrate that the algorithm is faster in adapting to sudden increasing levels of noise than existing algorithms and suitable for real-time implementation