leooo

Result of retrain is not good as pretrain models provided.

14

Thanks for your wonderful job. @breizhn I use this project to retrain on DNS-challenge dataset that was updated recently. The denoised results of the retraining model is a little worse...

Confusion about computation in paper 'Simple Recurrent Units for Highly Parallelizable Recurrence' ?

4

Thanks for your job. @taoleicn @taolei87 In the paper 'Simple Recurrent Units for Highly Parallelizable Recurrence', I found the following computation:  >To facilitate parallelization, our light recurrence component uses...

Thanks for your job. I find your nnwpe method outperforms other traditional methods on wer results of asr system after feeding data processed by wpe method or none. Is your...

About 'min_max' and 'kld' quantization

6

I check function 'generate_model' carefully, but still confused about 'min_max' quantization approach. I also look up the related [documents](https://majianjia.github.io/nnom/api_nnom/) and do not find the answers. For example, why 'find_dec_bits_max_min' function...

Results of mfcc in main.c is different from that in python with the same setting.

13

### Results of mfcc in main.c is different from that in python with the same setting. I give the same input(512 zero samples.) to the 'mfcc‘ api in main.c and...

Covert onnx lstm model to paddlelite model failed

2

我利用tf.keras训练两层lstm+一层全连接降噪模型，并利用keras2onnx转为onnx模型，然后计划将onnx lstm 转换为paddlelite model, 但转换失败。我也尝试利用onnx-simplifier来简化onnx lstm model， 简化成功后转换paddlelite model [仍然失败，这一步参考。](https://github.com/PaddlePaddle/X2Paddle/issues/511) ## 示例 模型来源： onnx model from keras2onnx 模型文件： 链接：https://pan.baidu.com/s/19z5Efr4siw1_ByecnH_4PA 提取码：kpj3 转换过程出错提示如下： onnx-simplifier 简化前：  onnx-simplifier 简化后：  期待任何回复，感谢。

Thanks for your great job. I have a problem with the training target. I do not know which I should take as the training target among nearend-speech with rir and...

Thanks for your great job. I checked your paper today, but still confused about some training details. It's no doubt this is a wonderful aec job. I am looking forward...

'config = { "add_AGC": 1}' error

1

` #Test AGC input_file = ["agc_testaudio3.wav"] out_file = ["agc_testaudio3_athenaout.wav"] # config = { "add_AGC": 1, "add_NS":0, 'add_AEC': 0} config = { "add_AGC": 1} athena_signal_process(input_file, out_file, config) ` I want to...

你好，非常谢谢你详细的工作。 在看完你的blog后，我有个问题，计算IRM的近端语音信号，是纯干净的近端语音，还是带噪声带混响的近端语音？ 我尝试使用纯干净的近端语音后没有得到好的loss结果。 感谢任何建议。 @LXP-Never