Add files via upload

Speech_Recognition/preprocessdata.py

@@ -0,0 +1,70 @@
+import tensorflow as tf
+import os
+import librosa
+import librosa.display
+import matplotlib.pyplot as plt
+import seaborn as sns
+import numpy as np
+import random
+import scipy.io as sc
+sns.set()
+#import IPython.display
+
+data_train_path='./dataset/training/'
+data_test_path='./dataset/test/'
+batchsize=100
+SAMPLINGRATE=16000
+PADDING=44
+MFCCSNR=20
+labels_map={'bed':1,'bird':2,'cat':3,'dog':4,'down':5,'eight':6,'five':7,'four':8,'go':9,'happy':10,'house':11,'left':12,'marvin':13,
+'nine':14,'no':15,'off':16,'on':17,'one':18,'right':19,'seven':20,'sheila':21,'six':22,'stop':23,'three':24,'tree':25,'two':26,'up':27,'wow':28,'yes':29,'zero':30}
+
+def one_hot_vector(label):
+    encoding=[0]*len(labels_map)
+    encoding[labels_map[label]-1]=1
+    return encoding
+
+def get_mfcc(path, padding=PADDING):
+    wave,sr=librosa.load(path,sr=SAMPLINGRATE,mono=True)
+    #librosa.display.waveplot(wave,sr=sr)
+    mfccs=librosa.feature.mfcc(y=wave,sr=SAMPLINGRATE,n_mfcc=MFCCSNR)
+    mfccs=np.pad(mfccs,((0,0),(0,padding-len(mfccs[0]))),mode='constant')
+    return mfccs
+
+def load_batch(size,path):
+    X=[] #mfccs for each audio
+    Y=[] #one hot vectors of labels
+    Z=[] #labelse
+    path=os.path.join(path, '*','*.wav')
+    waves=tf.gfile.Glob(path) #all files in subdirectories
+    allFiles=len(waves)
+    print("Number of files being processed: ",allFiles)
+    #waves=waves[0:size]
+    for wave_path in waves:
+        _,label=os.path.split(os.path.dirname(wave_path))
+        X.append(get_mfcc(wave_path))
+        Y.append(one_hot_vector(label))
+        Z.append(labels_map[label])
+    return X,Y,Z
+
+def load_data(mode):
+    if mode=='train':
+        path=data_train_path
+        saveFile='train20.mat'
+    elif mode=='test':
+        path=data_test_path
+        saveFile='test20.mat'
+    else:
+        saveFile=''
+    print("Processing data in ", path)
+    X,Y,Z=load_batch(10,path)
+    arrayformx=np.array(X)
+    arrayformy=np.array(Y)
+    try:
+        sc.savemat(saveFile, dict(mfccs=arrayformx, onehot=arrayformy, labels=Z))
+    except:
+        sc.savemat('some20.mat', dict(mfccs=arrayformx, onehot=arrayformy, labels=Z))
+    print("Saved processed data in", saveFile)
+
+
+load_data('train')