基于BLSTM神经网络的回声和噪声抑制算法

考虑到非线性回声和非平稳噪声对智能设备回声消除算法的影响,论文提出一种基于双向长短时记忆(Bidirectional Long Short-Term Memory,BLSTM)神经网络的回声和噪声抑制算法。该算法首先采用多目标预处理模型,同步估计出回声和噪声信号的幅度谱;然后将其作为回声和噪声抑制模型的输入特征,进而估计出目标语音信号的理想比例掩模;最后通过联合训练两个模型得到最优回声和噪声抑制模型。实验结果表明,在非线性回声和非平稳噪声的环境下,该算法均取得了较好的回声和噪声抑制效果,语音失真较小。

Echo and noise suppression algorithm based on BLSTM Neural Network

Considering the influence of nonlinear echo and non-stationary noise on the echo cancellation algorithm of intelligent equipment, this paper proposes an echo and noise suppression algorithm based on bidirectional long short-term memory neural network. Firstly, the multi-target preprocessing model is used to estimate the amplitude spectrum of echo and noise signals synchronously. Then it is used as the input feature of echo and noise suppression model to estimate the ideal ratio mask of target speech signal. Finally, the optimal echo and noise suppression models are obtained through the joint training of the two models. The experimental results show that the proposed algorithm has better echo and noise suppression effects and less speech distortion in the environment of nonlinear echo and non-stationary noise.