ProWav

You can use this for preprocessing wave files.

Usage

Please install prowav by using pip.

pip install prowav  

Usage

ProWav can calculate mfcc spectrogram and pad for batch execution

from prowav import ProWav  

prowav = ProWav(["path/to/wave/data_1.wav", "path/to/wave/data_2.wav"])  

frame_width = 20 # the length of a frame (ms)
stride_width = 20 # the frame interval (ms)
n_mfcc = 26 # the number of features by mfcc features (If you want to use mfcc preprocessing, you should specify this value)  
mode = 'MFCC'
window_func = 'hamming' # the name for window function
data = prowav.prepro(frame_width=frame_width,stride_width=stride_width,mode=mode,
                                       n_mfcc=n_mfcc, window_func=window_func)
# >> (num_files, num_frames, n_mfcc)    

If you want use fft spectrogram, please specify the mode, "fft".

prowav = ProWav(["path/to/wave/data_1.wav", "path/to/wave/data_2.wav"])
frame_width = 20
stride_width = 20
mode='fft'
window_func='hamming'  
data = prowav.prepro(frame_width=20,stride_width=20,\
      mode=mode, window_func=window_func)  
# >> (num_files, num_frames, num_features)

You can also use mel-spectrogram. Specify the mode, "mel_spec"

prowav = ProWav(["path/to/wave/data_1.wav", "path/to/wave/data_2.wav"])
frame_width = 50 
stride_width = 50 
mode='MFCC'
n_mfcc = 26
window_func='hamming'
data = prowav.prepro(frame_widh=frame_width, stride_width=stride_width,
                  mode=mode, window_func=window_func, n_mfcc=n_mfcc)

You can use zero-padding or repeat-padding.

prowav = ProWav(["path/to/wave/data_1.wav", "path/to/wave/data_2.wav"])
frame_width = 50 
stride_width = 50 
mode='mel_spec'
n_mels = 50
window_func='hamming'
data_zero = prowav.prepro(frame_widh=frame_width, stride_width=stride_width,
                  mode=mode, n_mels=n_mels,window_func=window_func, zero_padding=True) # zero padding 
data_repeat = prowav.prepro(frame_widh=frame_width, stride_width=stride_width,
                  mode=mode, ne_mels=n_mels, window_func=window_func, repeat_padding=True) # repeat padding 

Just loading wave data is possible.

prowav = ProWav(["path/to/wave/data_1.wav", "path/to/wave/data_2.wav"]) 

prowav.load_wav() # loading wav file into this class.

prowav.data # the list of ndarray. Raw data are listed.

You can choose parallel option.

prowav.load_wav(parallel=True)

prowav.load_wav(parallel=True, verbose=5)