def files_to_data(files, file_labels, threshold):
uniqueLabels = np.unique(file_labels)
labelString2Int = {s:i for i,s in enumerate(uniqueLabels)}
labels=[]
data=[]
for file,file_label in zip(files,file_labels):
wav = segment.wav_to_np(file)[:,0]/32768. # left CH
chunks = segment.chop_all(wav, threshold, afterlength=700, prelength=0)
chunks = map(transform.sndFeature, chunks)
data += list(chunks)
labels += [labelString2Int[file_label],]*len(chunks)
print 'Chunked {} examples for {}'.format(len(chunks),file)
X = np.array(data)
y = np.array(labels)
return X,y,uniqueLabels
from sklearn.externals import joblib
import sys
tap_recog = joblib.load('forest_recog.bin')
p = pyaudio.PyAudio()
realtime = True
over=False
if realtime:
stream = p.open(format=p.get_format_from_width(2),
channels=1,
rate=44100,
input=True)
else:
chunk = segment.wav_to_np('snaps/gss.wav')[:,0]/32768.
# segment.play_wav(chunk)
frames = []
CHUNK = 10000
threshold = 0.4
text=[]
try:
last_click = None
while True:
print 'loop'
if realtime:
chunk = stream.read(CHUNK)
chunk = segment.decode(chunk, 1)[:,0]/32678.
elif over:
break
