关键词: 鼾声, 降噪, 谱减法, 维纳滤波法, 子空间投影法, 信噪比 

Abstract:  A method to improve the signalto-noise ratio (SNＲ) of snoring sound is proposed in combination with the traditional spectral subtraction and Wiener filter in this study． Firstly， the noisy snoring signals are slightly enhanced by traditional spectral subtraction by projecting the noisy snoring to noise space and signal space by a method of subspace projection so that the SNＲ is obtained． Then a transfer function of Wiener filter from SNＲ is obtained． Finally，when the snoring sound processed by traditional spectral subtraction is filtered by the Wiener filter， the noise in snoring sound can be further reduced． The results of the simulation of snoring sound with additive white noise show that the method used in the study gets a higher SNＲ than that by the traditional subtraction or Wiener filter． It is proved that the method can do better than the one of traditional subtraction or Wiener filter in noise reduction． 

Key words: snoring sound, noise reduction, spectral subtraction, Wiener filter, subspace projection, signaltonoise ratio