卷神经网络用于声纹识别

来自github，不可用于商业用途，仅供学习。

function [ cost grad numTotal pred_cell ] = drdae_obj（ theta eI data_cell targets_cell fprop_only pred_out）
%PRNN_OBJ MinFunc style objective for Deep Recurrent Denoising Autoencoder
%   theta is the full parameter vector
%   eI contains experiment / network architecture
%   data_cell is a cell array of matrices. Each a distinct length is a cell
%             entry. Each matrix has a time series example in each column
%   targets_cell is parallel to data but contains the labels for each time
%   fprop_only is a flag that only computes the cost no gradient
%   eI.recurrentOnly is a flag for computing gradients for only the
%   recurrent layer. All other gradients set to 0
%   numTotal is total number of frames evaluated
%   pred_out is a binary flag for whether pred_cel

 属性            大小     日期    时间   名称
----------- ---------  ---------- -----  ----
     目录           0  2014-06-06 01:27  rnn-speech-denoising-master\
     文件          23  2014-06-06 01:27  rnn-speech-denoising-master\.gitignore
     文件        1580  2014-06-06 01:27  rnn-speech-denoising-master\README.md
     文件        8337  2014-06-06 01:27  rnn-speech-denoising-master\drdae_obj.m
     文件         912  2014-06-06 01:27  rnn-speech-denoising-master\getAllFiles.m
     文件        1567  2014-06-06 01:27  rnn-speech-denoising-master\htkread.m
     文件        2469  2014-06-06 01:27  rnn-speech-denoising-master\htkwrite.m
     文件        2698  2014-06-06 01:27  rnn-speech-denoising-master\initialize_weights.m
     文件        8646  2014-06-06 01:27  rnn-speech-denoising-master\load_aurora.m
     文件        2363  2014-06-06 01:27  rnn-speech-denoising-master\rnn_params2stack.m
     文件        2305  2014-06-06 01:27  rnn-speech-denoising-master\rnn_stack2params.m
     文件         367  2014-06-06 01:27  rnn-speech-denoising-master\save_callback.m
     文件        2633  2014-06-06 01:27  rnn-speech-denoising-master\train_aurora_local.m
     文件        1570  2014-06-06 01:27  rnn-speech-denoising-master\write_rec_dataset.m