Silence_Trim.py

'''
This script requires three libraries: librosa, numpy, scipy
If there is an error, please setup the libraries.
To install, run the following command in the terminal.

sudo pip install numpy, scipy, librosa

'''
import librosa, os
import numpy as np
import scipy.io

frequency = 22050  #The frequency of wave file
maxv = np.iinfo(np.int16).max   #32767
wav_Folder = "D:/Simulation_Raw_Data/EARS/SAY_VOICES_TIMIT_LEX"    #Change the folder that there is the wave files


for root, directory_List, file_Name_List in os.walk(wav_Folder):
    if not os.path.exists(root.replace(wav_Folder, wav_Folder + "_Trim")):  #When there is no export folder, make new one.
        os.makedirs(root.replace(wav_Folder, wav_Folder + "_Trim"))
    for file_Name in file_Name_List:
        if not file_Name.lower().endswith('.wav'):  #Only wav file is trimmed.
            continue
        sig = librosa.core.load(os.path.join(root, file_Name).replace("\\", "/"), sr = frequency)[0] #Signal load
        trimmed_Sig = librosa.effects.trim(sig, frame_length = 32, hop_length=16)[0]    #Delete the silence of front and back
        scipy.io.wavfile.write(os.path.join(root.replace(wav_Folder, wav_Folder + "_Trim"), file_Name).replace("\\", "/"), rate = frequency, data=(trimmed_Sig * maxv).astype(np.int16))  #Signal save