A Method Based on Double-channel Adaptive Noise Cancelling for Speech Enhancement

By using the noisy correlations between the main and referenced channels in adaptive noise cancelling (ANC),a new adaptive filtering algorithm is proposed for speech enhancement.In an additive white Gaussian noise environment with low signal-to-noise ratio (SNR) ranging from 0 dB to-10dB,speech enhancement is researched on the case of the reference input noise mixed with crosstalk whether or not.The result can be described as follows:Assuming that the reference input containing crosstalk,the improved SNR and PESQ score by the new algorithm is more about 14dB and 1.13 than those of NLMS algorithm,respectively.Without the above assump- tion,the improved SNR and PESQ score become 5dB and 0.83,respectively.Moreover,the listening quality of the enhanced speech is remarkably improved.