ID 原文 译文
23945 虽然其较传统的声学建模方法有很大优势，但相对较高的计算复杂度限制了这种神经网络的应用，特别是在实时应用场景中。 Although RNNs show many advantages over traditional acoustic modeling methods, the inherent higher computational cost limits its usage, especially in real-time applications.
23946 由于递归神经网络采用的输入特征通常有较长的上下文，因此利用重叠信息来同时降低声学后验和令牌传递的时间复杂度成为可能。 Noticing that the features used by RNNs usually have relatively long acoustic contexts, it is possible to lower the computational complexity of both posterior calculation and token passing process with overlapped information.
23947 该文介绍了一种新的解码器结构，通过有规律抛弃存在重叠的帧来获得解码过程中的计算开销降低。 This paper introduces a novel decoder structure that drops the overlapped acoustic frames regularly, which leads to a significant computational cost reduction in the decoding process.
23948 特别地，这种方法可以直接用于原始的递归神经网络模型，只需对隐马尔可夫模型(Hidden Markov Model, HMM)结构做小的变动，这使得这种方法具有很高的灵活性。 Especially, the new approach can directly use the original RNNs with minor modifications on the HMM topology, which makes it flexible.
23949 该文以时延神经网络为例验证了所提出的方法，证明该方法能够在精度损失相对较小的情况下取得 2~4 倍的加速比。 In experiments on conversation telephone speech datasets, this approach achieves 2 to 4 times speedup with little relative accuracy reduction.
23950 该文提出一种低复杂度的迭代大数逻辑 LDPC 译码算法，在迭代过程中所有的译码信息都以二元形式进行传递、处理和迭代更新。 A low complexity iterative majority-logic decoding algorithm is presented. For the presented algorithm, binary decoding messages are involved in the message passing, processing and updating process.
23951 所提算法不需要计算外信息，而是利用 Tanner 图上伴随式的对错状态来评判节点可靠度。 Instead of computing the extrinsic information, the presented algorithm computes the reliability measure based on syndrome states (correct or error) in the Tanner graph.
23952 与现有的几种迭代大数逻辑译码算法相比，该文算法也不需要信息修正处理，避免了相应的实数乘法操作，具有很低的译码复杂度。 Compared with several existing iterative majority-logic decoding algorithms, the presented algorithm does not require the information scaling and hence can avoid the corresponding real multiplication operations. This leads to very low decoding complexity.
23953 此外，该文引入一种特殊的量化处理函数，并给出了基于离散密度进化的参数优化过程。 Furthermore, a special quantization is combined with the presented algorithm. The optimization method is also given based on the discrete Density Evolution (DE).
23954 实验仿真表明，该文所提算法与原算法相比，在 AWGN 信道下可获得约 0.3~0.4 dB 的性能提升。 Simulation results show that, compared with the original algorithm, the presented algorithm can achieve about 0.3~0.4 dB performance gain over the Additive White Gaussian Noise (AWGN) channel.