FUNCTION OF BUCKET ITERATOR about speech-emotion-recognition

[amanbasu]

In this project, we have used human speech signals for detecting the emotions. Now since these voice samples can vary in time, we need to convert them to a common shape that requires padding (neural networks require data in a fixed shape for training).
Either whole data can be padded altogether or batch-wise, the later one is more memory efficient and we have used Bucket Iterator for the same.
Bucket Iterator takes the batch size as the parameter and returns a random padded batch of data including both input and its corresponding label.

[Jon-Zbw]

I want to know what is the main function of 'Mel.py' , Just want to show the difference between FBANK and MFCC?

[amanbasu]

Yes, it is just for reference and understanding Mel.

[Jon-Zbw]

when i copy your ‘class BucketedDataIterator（）：’
I don't know where the problem is. Can you help me?
Traceback (most recent call last):
File "D:/Project/pycharm/RNN/create_mfcc.py", line 164, in
mini_batch = tr.next_batch(128)
File "D:/Project/pycharm/RNN/create_mfcc.py", line 160, in next_batch
x_i[:res['lengths'].values[i]] = res['MFCC'].values[i]
TypeError: slice indices must be integers or None or have an index method

[Jon-Zbw]

my file like this：
........................................................... MFCC ........................................... lengths
0 [[-10.494777560823318, 19.205118287686812, 16.... ... 1099.0
1 [[-11.632857909892738, 5.8760442872698695, 29.... ... 1099.0
2 [[-8.773702814090234, -12.86215112148708, -7.2... ... 299.0
3 [[-8.411895008114222, -1.9307392294278005, -7.... ... 299.0
4 [[-6.969186995608002, -17.02179303376442, -23.... ... 299.0