Rong Xiaobin
Rong Xiaobin
这些是我从 DNS3 blind test set 中挑选的测试样例，希望对你有帮助。 [test examples.zip](https://github.com/Xiaobin-Rong/deepvqe/files/13373892/test.examples.zip)
I'm not sure if what you mentioned refers to the sentence in the paper: "The first microphone encoding block has 64 filters and all the following microphone encoding blocks have...
Thank you for your attention and praise for my code. As I mentioned in the README, my focus in reproducing DeepVQE was solely on its speech enhancement performance, so the...
You're welcome! I'm glad you found the training template useful. I really appreciate your kind words. Thanks again for your support!
The only difference between the two lies in the implementation of the CCM module, where deepvqe utilizes real number multiplication exclusively, while deepvqe_v1 uses complex number multiplication. I believe that...
RNNoise的 MACs，使用的是 [DeepFilterNet2](https://arxiv.org/pdf/2205.05474) 的 Table 1 中提供的数据。
你说得对，RNNoise原文把 FFT 和 IFFT 也算进运算量里了。这样我的对比确实存在不严谨的地方，感谢指出。但 RNNoise 只是 GTCRN 的对比对象之一，GTCRN 在 VCTK-DEMAND 数据集上的 PESQ=2.87 是客观成立的。
The paper is currently not available for public access online. According to the official schedule provided by ICASSP 2024, the paper will be made available for online download starting from...
@shenbuguanni 论文同时与专利申请有关，所以暂不方便发布预印版。
@shenbuguanni FLOPs 与 MACs 本来不是一个东西，MACs 理论上是 FLOPs 的两倍，现在大家常常把这两者弄混淆。按照我的计算，get_model_complexity_info 输出的应该是 FLOPs。