EaBNet
EaBNet copied to clipboard
This is the repo of the manuscript "Embedding and Beamforming: All-Neural Causal Beamformer for Multichannel Speech Enhancement", which was submitted to ICASSP2022.
您好,您能给一下EaBNet在8k采样率下的配置吗? 我修改了这里的`sr=8000`和`fft_num=160`,结果会报错 https://github.com/Andong-Li-speech/EaBNet/blob/719f764bbc5128df8cf5eaf59d91e7a2a577939d/EaBNet.py#L768C1-L772C60
你好,我在处理您论文中提到的数据,其中RIR参数中阵列位置没有指明,请问您的阵列的location在哪里呢? 此外,在我把阵列位置设定为房间正中间后,仿真得到的RIR参数后,干净语音clean和rir1卷积,噪声信号noise和rir2卷积,并得到多通道信号,相加之后得到含噪信号noisy,请问这个noisy的SDR是否与SNR接近, 文章table2中:Noisy - - - - - 1.45/29.62/-4.89 1.62/37.71/-1.93 1.74/41.85/0.05 1.87/49.42/2.05 1.67/39.65/-1.18。分别接近(-5dB,-2dB,0dB,2dB) 但是我的仿真结果SDR与SNR在高信噪比时是不接近的,并且stoi也比较大
Thanks for your code. I wonder how to convert from enhanced stft generated from the generator to the waveform so that I can hear the enhanced audio I know that...
您好,看了您的论文《EMBEDDING AND BEAMFORMING: ALL-NEURAL CAUSAL BEAMFORMER FOR MULTICHANNEL SPEECH ENHANCEMENT》和对应的代码,有一个疑问:多通道的target是什么?看论文和代码都没有具体说,您是以某一个通道作为整体的target还是多个通道有多个target,然后分别进行处理的?