def main():
if len(argv) != 2:
print ("Need weights file.")
return
weights_file = path.join(CLASSIFIERS_DIR,argv[1])
fcall = "SMILExtractPA -C %s -sampleRate 44100 -channels 1 -O %s -noconsoleoutput"\
%(path.join(CONF_BASE,CONF_RECORD_FILE),WAVE_FILE)
fcall = fcall.split(' ')
#record audio
print "\n\nPress any key to start recording...end recording with CTRL-C in new window."
raw_input()
rec_proc = sp.Popen(fcall,creationflags=(sp.CREATE_NEW_CONSOLE))
rec_proc.wait()
#get the features
X = np.array(parse_features(WAVE_FILE,SMILE_CALL,CONF_FILE))
X = np.reshape(X,(1,X.shape[0]))
#run the classifier
N = NN()
N.load(weights_file)
prediction = N.predict(X)[0]
#get language and print result
language = LANGS[prediction]
print "That sounds like %s to me." % language
def main():
#set up argument parsers
parser = argparse.ArgumentParser("Extract audio features from directory of files using SMILE.")
parser.add_argument('directory',metavar='DIRECTORY',type=str,help='directory to parse, relative to audio folder.')
parser.add_argument('-L',metavar="LANGUAGE",type=str,choices=LANGS,required=True,
help='one of ' + str(LANGS) + " add languages in Detector/configuration.py")
parser.add_argument('-o',metavar="OUTFILE",type=str,required=True,\
help ='location of output file relative to features folder, appends file if exists else creates it.')
parser.add_argument('-c',metavar="FEATURE CONFIGURATION FILE",type=str,required=False,default=CONF_FILE,\
help="configuration file to use defaults to CONF_FILE.")
parser.add_argument('--CALL',metavar="SMILE CALL",type=str,required=False,default=SMILE_CALL,\
help="call to SMILExtract defaults to SMILE_CALL")
args = parser.parse_args()
directory = path.join(AUDIO_DIR,args.directory)
lang = args.L
label = LANGS.index(lang)
outfilename = path.join(FEATURES_DIR,args.o)
configfile = args.c
smile_call = args.CALL
try:
data = loadmat(outfilename)
X = data['X'].tolist()
y = [x[0] for x in data['y'].tolist()]
except IOError:
X = []
y = []
for fname in listdir(directory):
features = parse_features(path.join(directory, fname),smile_call,configfile)
X.append(features)
y.append(label)
nLabels = len(LANGS)
results = {X_KEY : np.array(X),
Y_KEY :y,
NLABELS_KEY : nLabels,
FEATURE_SET_KEY : configfile,
LANGS_KEY : '-'.join(LANGS)}
savemat(outfilename,results)
