To address the limitations of the traditional wavelet threshold denoising function, an improved wavelet threshold function was proposed. The improved threshold function not only has good continuity but also overcomes the lack of the non-negative dead zone threshold function and considers the characteristic of the attenuation of the noise wavelet modulus values. In addition, the use of spectral flatness function corrects the threshold values adaptively. The simulation results have showed that the improved wavelet threshold can eliminate ground noise effectively, maintain higher speech quality and definition while improving the signal to noise ratio (SNR) of the output.