一种基于改进的量子神经网络的语音降噪方法

利用一种改进的量子神经网络（IPSO-QNN）在时域上对含噪的语音信号进行降噪处理，文中重点改进了QNN所涉及到的学习算法。针对粒子群算法本身存在早熟的不足，提出了一种改进的粒子群优化算法（IPSO）。该算法通过对早熟粒子的速度和位置叠加随机数据，使其离开局部最优，从而使算法具有更强的寻优能力。利用IPSO对量子神经网络的参数进行训练和学习，建立了比较高效的基于改进的量子神经网的语音信号滤波器，并通过Matlab软件建立实验平台，实验结果表明，新算法充分利用了量子神经计算的快速性以及粒子群算法的全局寻优能力，从而使该语音信号滤波器，具有良好的降噪性能。

A Method to Decrease Noise in Speech Based on Improved Quantum Neural Networks

In this paper, the IPSO-QNN is adopted to decrease noise of noisy speech signal, and focused on improving the learning algorithm of QNN. Aimed at PSO’s defect of prematurity, an improved particle swarm optimization (IPSO) is presented. The new arithmetic has better optimization performance by adding random data to premature particles’ speed and position. It was applied to parameter learning and training of QNN, a higher efficiency speech signal filter which based on IPSO-QNN was established. Experimental results of MATLAB simulation showed that the new arithmetic did a better job in decreasing noise as a filter of speech signal which make the best of faster quantum neural computation and PSO’s global optimization ability.