yong xu @ seattle

ReLU can be added into DevFunc.cu: ----------------------------------------------------------------------------------- ////////ReLU __global__ void kernSigmoid(int n, float* in_vec, float* out_vec) { int i = (blockIdx.x * blockDim.x) + threadIdx.x; if (i < n) //sigmoid...

The ReLU has already been updated in the code now.

you can use 'help/step1_DNNenh_for16kHz.m' instead of 'step1_DNNenh_for16kHz.exe', to see clearly. 'WAV2RAW.exe' also in help dir. regards, yong

copy step1_DNNenh_for16kHz.m to the father dir (same place with step1_DNNenh_for16kHz.exe), it should run ok with matlab. My OS is win10. WAV2RAW.exe is just there. copy is the OS command.

wav2raw. c source code: #define _CRT_SECURE_NO_DEPRECATE /*----------------- - File Inclusions _-----------------_/ #include #include #include #include extern int main (int argc, char *argv[]) { FILE *fp_in, *fp_out; ``` short data; if(3...

Hi, the initialized model was not pre-trained. Just with random initialization.

ok, it is here: https://github.com/yongxuUSTC/DNN-for-speech-enhancement/tree/master/toolbox/weights source code for initializing your model weights randomly and change back the weights for matlab decoding

Yes, correct. Just with fine-tuning process with random initialization. I once tried RBM-based pre-training which did not work.

The direct mapping is from noisy speech log-power spectra to clean speech log-power spectra. Additionally, you can also predict noise log-power spectra, ideal binary mask, or ideal ratio mask to...

where do you find "“timit_aurora4_115NT_7SNRS_each190_80uuts_noisy_lsp_be_random_linux_global_mv.mat”" ? I think i use a different one: https://drive.google.com/file/d/0B5r5bvRpQ5DRR1lIV1hpZ0RLQ0E/view