ID 原文 译文
39266 比GOP方法的诊断准确率相对提升15.5%，并且模型相较于无标注经验汉语母语者能检测出更多偏误。 The accuracy rate of cERN is 76.6%, which improve 15.5% than GOP's accuracy rate and is better than the result of untrained annotates.
39267 传统的线性主动噪声控制算法在噪声信号或噪声通道呈现非线性特性的情况下控制效果不佳。 The performance of the traditional linear active noise control algorithms degrades when the noise signal or primary path is nonlinear.
39268 核-滤波最小均方误差算法(Kernel Filtered x Least Mean Square，KFxLMS)通过将输入噪声信号映射到高维再生核希尔伯特空间，再用线性方法在高维空间中进行处理。 The Kernel Filtered x Least Mean Square(KFxLMS) algorithm maps the input noise signal to the higher-dimensional reproducing kernel Hilbert space, and then adopts the linear method to process the mapped signal.
39269 然而，随着新噪声信号的输入，KFxLMS算法递增的核函数运算需要较高的成本。 However, with the feed of new noise signal, the KFxLMS algorithm requires a high cost to realize the kernel calculation.
39270 为进一步改进KFxLMS算法，本文提出了随机傅里叶特征核滤波最小均方误差算法(Random Fourier Feature-Kernel Filtered x Least Mean Square，RFF-KFxLMS)。 In this paper, a nonlinear active noise control algorithm Random Fourier Feature-Kernel Filtered x Least Mean Square(RFF-KFxLMS) algorithm is proposed.
39271 在仿真实验部分讨论了算法的参数选择，给出了算法的计算耗时，并验证了提出的RFF-KFxLMS算法在非线性噪声通道情况下，针对不同频率分量的正弦噪声都能够达到理想的性能。 In the simulation experiment, the parameter selection is discussed, and the consuming time of the algorithm is given. In the case of nonlinear primary path, the proposed RFF-KFxLMS algorithm is verified by comparative experiments to achieve ideal performance in condition of sinusoidal noises with different frequency components.
39272 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。 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.
39273 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; Firstly, the multi-target preprocessing model is used to estimate the amplitude spectrum of echo and noise signals synchronously.
39274 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; Then it is used as the input feature of echo and noise suppression model to estimate the ideal ratio mask of target speech signal.
39275 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。 Finally, the optimal echo and noise suppression models are obtained through the joint training of the two models.