【摘要】 描述了在诸如封闭环境的散射噪声场中,一种基于ADALINE神经网络的麦克风阵列语音增强方法。该方法充分利用了波束形成和神经网络的优点,对基于MLP-BP结构进行了改进,以提高神经网络的学习方法的收敛速度。仿真结果表明,和MLP-BP相比,该方法具有更好的噪声抑制性能。更多还原

【Abstract】 Adaptive Linear Neuron (ADALINE) neural network based on microphone array speech enhancement method was described, concerning with noise scattering field in closed environment, which take advantage of characteristics of both beamforming and neural network. An improvement was made upon the structune on basis of MLP-BP, so as to enhance convergent speed of neural network learning method. Simulation results indicate that compared with Multiplier Linear Perception neural network using backpropagation (MLP-BP) algorithm, the proposed method exhibits higher processing speed and better noise suppression performanc更多还原