import os
import random

path_timit = '/Users/musy/code/TIMIT'
ending = 'WAV'

speakers = {}
for root, dirs, files in os.walk(path_timit, topdown=False):
    allowed_files = (filename for filename in files if ending in filename)
    for f in allowed_files:
        speaker_name = root.split('/')[-1]
        if speaker_name in speakers:
            speakers.get(speaker_name).get('sound_files').append(os.path.join(root, f))
        else:
            speakers[speaker_name] = {'sound_files': [os.path.join(root, f)]}


def select_random_speaker(sex=None):
    if sex is None:
        name = speakers.keys()[random.randint(1, 630)]
        return name, speakers[name]


speaker = select_random_speaker()
speaker

%pylab inline
import wave
import sys

wav = speaker[1].get('sound_files')[10]
#wav = '/Users/musy/Desktop/SA1.WAV'

spf = wave.open(wav,'r')
print spf.getframerate(), spf.getnchannels(), spf.getnframes(), spf.getsampwidth(), spf.getcomptype(), spf.getcompname()

spf = wave.open(wav,'r')

#Extract Raw Audio from Wav File
signal = spf.readframes(-1)
signal = np.fromstring(signal, 'Int16') #data in signal


#If Stereo
if spf.getnchannels() == 2:
    print 'Just mono files'
    sys.exit(0)

plt.figure(1)
plt.title('Signal Wave...')
plt.plot(signal)

from IPython.display import Audio
Audio(wav)

from scipy.io import wavfile

fpaths = '/Users/musy/Desktop/SA1.WAV'

fs, d = wavfile.read(fpaths)

fs, d.shape

Audio(d, rate=fs)

signal

signal.shape

import librosa

d.shape

mfccs = librosa.feature.mfcc(d, sr=fs, n_mfcc=13)
mfccs

from librosa.feature import melspectrogram
from librosa.display import specshow
S = melspectrogram(d, sr=fs, n_fft=1024)
logS = librosa.logamplitude(S)
specshow(logS, sr=fs, x_axis='time', y_axis='mel')