def do_GET(s):
s.send_response(200)
s.send_header("Content-type", "text/html")
s.end_headers()
path = s.path
if '-' in path:
acao, nome, pessoa = path.split('-')
if acao.replace('/', '') == 'gravar':
#Adicionando pessoa ao banco
db.add_model((caminhoVoz + nome), pessoa)
v = Voiceid(db, (caminhoVoz + nome + ".wav"))
s.wfile.write("Gravado")
if acao.replace('/', '') == 'biometria':
#Verificação de arquivo de voz
v = Voiceid(db, (caminhoVoz + nome + ".wav"))
usuario = v.extract_speakers()
s.wfile.write("%s" % usuario)
if not "S0 (unknown)" in str(usuario):
ser.write(ledLigado)
for the_file in os.listdir(caminhoVoz):
file_path = os.path.join(caminhoVoz, the_file)
try:
if os.path.isfile(file_path):
os.unlink(file_path)
elif os.path.isdir(file_path):
shutil.rmtree(file_path)
except Exception, e:
print e
def main():
parser = argparse.ArgumentParser(version="1.0", prog="Voice Recognition")
parser.add_argument("-a", "--add", action="store_true", help="Add a new source")
parser.add_argument("audio", type=str, help="Name of the audio file (*.wav)")
args = parser.parse_args()
audio = HOME_FOLDER + args.audio
db = GMMVoiceDB(HOME_FOLDER)
if args.add:
name = str(time.strftime("%d%m%Y%H%M%S"))
db.add_model(audio.split(".")[0], name)
else:
v = Voiceid(db, audio)
v.diarization()
dic = db.get_speakers()
dist = []
d2u = {}
for gender in dic:
users = dic[gender]
for u in users:
dist.append(db.match_voice(audio, u, gender)[u])
d2u[dist[-1]] = u
dist.sort(None, None, True)
if (dist[0] - dist[1] + CEIL) > EPS:
return d2u[dist[0]]
else:
return "unkonwn"
def bestSpeaker(file):
"""Returns best matching speaker: String in voice db for file: String. """
v = Voiceid(db, file, True)
v.extract_speakers()
clu = v.get_clusters().values()[0]
if len(clu.get_best_five()) <= 0:
return None;
return clu.get_best_five()[0][0]
def voicerec(fname):
db = GMMVoiceDB('models')
v = Voiceid(db,fname)
v.extract_speakers()
txt=''
for c in v.get_clusters():
cl= v.get_cluster(c)
cluster=str(cl)
cluster=cluster.split(' ')
cluster=cluster[1]
cluster=cluster.strip('(')
cluster=cluster.strip(')')
txt=txt+'Speaker : '+cluster
seg=str(cl.print_segments())
txt=txt+'\n'+seg
out.insertPlainText(txt+"\n")
def set_name(device_id, voice_path, new_name):
print "set " + voice_path + " to: " + new_name
new_name = new_name.replace(' ', '')
if not new_name.isalnum():
print 'error: SPEAKER_ID must be alphanumeric'
return
if new_name in db.get_speakers()['U'] or new_name in db.get_speakers(
)['M'] or new_name in db.get_speakers()['F']:
voice = Voiceid(db, voice_path, single=True)
voice.extract_speakers(quiet=True, thrd_n=3)
cluster = voice.get_cluster('S0')
cluster.set_speaker(new_name)
voice.update_db()
return
try:
# assume only one speaker in one sample
ww = fm.file2wav(voice_path)
file_basename, extension = os.path.splitext(ww)
db.add_model(file_basename, new_name)
os.remove(file_basename + ".seg")
os.remove(file_basename + ".ident.seg")
os.remove(file_basename + ".init.gmm")
except IOError:
print "voice file doesn't exist"
except OSError:
print "WARNING: error deleting some intermediate files"
except TypeError:
print "Type error"
def recognize(device_id, voice_path):
# voice_db_lock.acquire()
print db.get_speakers()
# assume only one speaker in one sample, To Do: multiple speakers in one sample
# set to True to force to avoid diarization, in case a single speaker in the file
try:
voice = Voiceid(db, voice_path, single=True)
# extract_speakers(interactive=False, quiet=False, thrd_n=1)
voice.extract_speakers(quiet=True, thrd_n=3)
# clusters = voice.get_clusters()
cluster = voice.get_cluster('S0')
# speaker = cluster.get_best_speaker()
speaker = "unknown"
speakers = cluster.get_best_five()
if len(speakers) > 0:
value = speakers[0][1]
if value > -33.0:
speaker = speakers[0][0]
# speaker = cluster.get_speaker()
print speaker
payload = {
'audio': 'http://52.24.205.33/voice/' + voice_path,
'userName': speaker,
'time': '1'
}
requests.post('http://129.236.234.21:8080/message',
data=json.dumps(payload),
headers=HEADERS)
client.publish("ais/recognize/result/" + device_id + "/" + voice_path,
speaker)
os.remove(voice.get_file_basename() + '.seg')
os.remove(voice.get_file_basename() + '.g.seg')
os.remove(voice.get_file_basename() + '.s.seg')
w = voice.get_file_basename() + '.wav'
if voice.get_filename() != w:
os.remove(w)
shutil.rmtree(voice.get_file_basename())
except IOError:
print "voice file doesn't exist"
def GetSpeakers(voicedb, audiofile):
from voiceid.sr import Voiceid
from voiceid.db import GMMVoiceDB
import logger
try:
v = Voiceid(voicedb, audiofile)
except:
logger.write("e", "File: voicehandler.py | Function: GetSpeakers | Error: Could not call Voiceid constructor!")
return False
try:
v.extract_speakers()
except:
logger.write("e", "File: voicehandler.py | Function: GetSpeakers | Error: Could not extract speakers from file!")
return False
try:
for c in v.get_clusters():
cluster = v.get_cluster(c)
voicedb.set_maxthreads(5)
except:
logger.write("e", "File: voicehandler.py | Function: GetSpeakers | Error: Could not get clusters!")
return False
return cluster
def main():
parser = argparse.ArgumentParser(version='1.0',prog='Voice Recognition')
parser.add_argument("-a", "--add", action="store_true", help="Add a new source")
parser.add_argument("audio", type=str, help="Name of the audio file (*.wav)")
args = parser.parse_args()
audio = HOME_FOLDER+args.audio
db = GMMVoiceDB(HOME_FOLDER)
if args.add:
print('Adding User...')
name = str(time.strftime("%d%m%Y%H%M%S"))
db.add_model(audio.split('.')[0], name)
print('User '+name+' Added!')
else:
print('Recognition...')
v = Voiceid(db, audio)
v.diarization()
dic = db.get_speakers()
dist = []
d2u = {}
for gender in dic:
users = dic[gender]
#print('Gender: ',gender)
for u in users:
#print(db.match_voice(audio, u, gender))
dist.append(db.match_voice(audio, u, gender)[u])
d2u[dist[-1]]=u
dist.sort(None, None, True)
if (dist[0]-dist[1]+CEIL)>EPS :
print(d2u[dist[0]])
else:
print('unkonwn')
print "Done!"
def segment_input(wavfile, dbpath='./voicedb'):
db = GMMVoiceDB(dbpath)
v = Voiceid(db, wavfile)
v.extract_speakers()
speaker_clusters={}
for c in v.get_clusters():
cluster = v.get_cluster(c)
print cluster
cluster.print_segments()
print
def voicerec(fname):
db = GMMVoiceDB('models')
v = Voiceid(db, fname)
v.extract_speakers()
txt = ''
for c in v.get_clusters():
cl = v.get_cluster(c)
cluster = str(cl)
cluster = cluster.split(' ')
cluster = cluster[1]
cluster = cluster.strip('(')
cluster = cluster.strip(')')
txt = txt + 'Speaker : ' + cluster
seg = str(cl.print_segments())
txt = txt + '\n' + seg
out.insertPlainText(txt + "\n")
if fnmatch.fnmatch(f, "*.wav") and f.find("##") > -1:
all_wav.append(f)
print all_wav
verify_goodness(all_wav)
print "verifica ok... procediamo"
logging.basicConfig(filename=DIR_PATH + "/result.log", level=logging.INFO)
db = GMMVoiceDB("/Users/labcontenuti/.voiceid/gmm_db")
filter_criteria = "Fabrizio_Gifuni##02Pasticciaccio.wav"
for wav in all_wav:
# if wav.find(filter_criteria)>-1:
print "--------------------" + wav + "------------------------"
v = Voiceid(db, DIR_PATH + wav)
v.extract_speakers()
logging.info("\n\n\n---------------------" + wav + "------------------")
wav_tostring = str(wav) + "; "
for c in v.get_clusters():
cluster = v.get_cluster(c)
for seg in cluster.get_segments():
print "\n seg: ", seg
print seg.get_start()
print seg.get_end()
print humanize_time(seg.get_duration())
print "cluster.get_duration ", cluster.get_duration()
logging.info("duration: " + str(cluster.get_duration()))
#Using voiceid(google opensource project) for voice recognisation using python
from voiceid.sr import Voiceid
from voiceid.db import GMMVoiceDB
db = GMMVoiceDB('mydir') #creating database/directory of registered voices
db.add_model('C:\Users\raman\Documents\raman.wma', 'Raman','M')
db.add_model('C:\Users\raman\Documents\lalit.wma', 'Lalit','M')
db.add_model('C:\Users\raman\Documents\aneesha.wma', 'Aneesha','F')
db.add_model('C:\Users\raman\Documents\priyanka.wma', 'Priyanka','F')
db.get_speakers() # this will get all the speakers in the database/directory
if(db.matchvoice('C:\Users\raman\Documents\lalit.wma', 'Lalit')) #the matchvoice function will compare the voices
{
print "Lalit is present" }
else print "Lalit is absent"
v = Voiceid(db)
v.extract_speakers()
for c in v.get_clusters():
cluster = v.get_cluster(c)
print cluster
cluster.print_segments()
def save_callback(self, file_=None):
""" Adds a file to the queue after it's been saved """
if self.test_mode == True:
vid = Voiceid(self.db, file_, single=True)
self.queue_processes.append((vid, None))
if options.DIR_PATH:
DIR_PATH=options.DIR_PATH
print "DIR_PATH ", DIR_PATH
if options.SCORE:
SCORE=float(options.SCORE)
print "SCORE ", SCORE
if options.SHELVE_DB:
SHELVE_DB=options.SHELVE_DB
print "SHELVE_DB ", SHELVE_DB
db = GMMVoiceDB(DB_PATH)
v = Voiceid(db, FILE_PATH)
#v.extract_speakers()
try:
v.extract_speakers_without_diarization()
print "Only speacker extraction"
except:
print "diarization and speacker extraction"
v.extract_speakers()
output = open(FILE_PATH+'.pkl', 'wb')
dic_pkl={}
for c in v.get_clusters():
cluster = v.get_cluster(c)
dic_pkl[cluster._label]=cluster
print "dic_pkl=", dic_pkl
pickle.dump(dic_pkl, output,-1)
output.close()
def recognition():
global audio, db, label
if audio != None:
print 'Comparing...'
#dic = db.get_speakers()
v = Voiceid(db, audio+'.wav')
#v.extract_speakers()
v.diarization()
dic = db.get_speakers()
dist = []
d2u = {}
for gender in dic:
users = dic[gender]
print('Gender: ',gender)
for u in users:
print(db.match_voice(audio+'.wav', u, gender))
dist.append(db.match_voice(audio+'.wav', u, gender)[u])
d2u[dist[-1]]=u
print('--------------')
print("Best Speaker (Own)")
dist.sort(None, None, True)
print( (dist[0]-dist[1]+CEIL),EPS,dist )
if (dist[0]-dist[1]+CEIL)>EPS :
print(d2u[dist[0]])
#print(dic)
print('--------------')
print("Best Speaker")
print("Audio: ",audio.split('/')[-1])
v.extract_speakers(False, True)
for c in v.get_clusters():
cluster = v.get_cluster(c)
print(cluster.get_best_speaker() )
print('------------')
'''
print 'Comparing...'
v = Voiceid(db, audio+'.wav')
#s = StringIO.StringIO()
#with s as sys.stdout:
v.extract_speakers(False, True)
#with open('output.txt','w') as sys.stdout:
for c in v.get_clusters():
cluster = v.get_cluster(c)
user = cluster.get_best_speaker()
print('------------')
if user != "unknown":
print("Best Speaker: %s" % (user) )
print("Speaker: %s" % (cluster.get_speaker()) )
print("Gender: %s" % (cluster.get_gender()) )
print("Distance: %s" % (cluster.get_distance()) )
print("Mean: %s" % (cluster.get_mean()) )
print("Dist. from mean: %s" % (cluster.get_m_distance()) )
print("Best five speakers: " % (cluster.get_best_five()) )
else:
print("User Unknown")
print("Gender: %s" % (cluster.get_gender()) )
print("Distance: %s" % (cluster.get_distance()) )
print("Best five speakers: " % (cluster.get_best_five()) )
print('------------')
#sys.stdout = sys.__stdout__
#labelB["text"] = "Users: " + str(listName)'''
print "Done!"
else :
print 'Record audio first!'
from voiceid.sr import Voiceid
from voiceid.db import GMMVoiceDB
# create voice db
db = GMMVoiceDB("voice")
# add models to db: params the basename of
# 'giov_wave.wav' and the speaker, Giovanni
# db.add_model("rec", "test")
# db.add_model("irina", "irina")
# print db.get_speakers()
# process a video/audio file containing various speakers
v = Voiceid(db, "rec.wav")
# extract the speakers from file and search into db
v.extract_speakers()
bestUsers = v.get_user()
print "Dict: " + str(bestUsers)
print "Max: " + max(bestUsers, key=bestUsers.get)
"""
for c in v.get_clusters():
cluster = v.get_cluster(c)
print cluster.get_best_speaker()
"""
def checkAudio(self, audio):
""" Check voices inside audio file """
v = Voiceid(self._db, audio)
v.extract_speakers(quiet=True)
return v
db = GMMVoiceDB('/Users/labcontenuti/.voiceid/gmm_db')
#print "adding"
#db.add_model('/Users/labcontenuti/Desktop/voiceid/adaltavoce/voci/PinoLoperfidoCap2MobyDick.wav', 'PinoLoperfido')
#print "added"
#v = Voiceid(db, '../Test_set_all/GianniVattimoPostmodernit-Lyotard.mp3')
#v = Voiceid(db, '../Test_set_all/Monica-Bellucci-interview.mp3')
#v = Voiceid(db, '../Test_set_all/Pasolini1971.mp3')
#v = Voiceid(db, '/home/felix/Desktop/u/facciamoiconti.mp3')
#v = Voiceid(db, '/home/felix/Desktop/u/facciamoiconti/S32.wav')
#v = Voiceid(db, '/Users/labcontenuti/Desktop/voiceid/adaltavoce/voci/FabrizioGifuni-promessisposi-p1.wav')
#v = Voiceid(db, '/Users/labcontenuti/Desktop/voiceid/adaltavoce/test/Piero_Baldini##mobyDick01_1sec.wav')
#v = Voiceid(db, '/Users/labcontenuti/Desktop/voiceid/adaltavoce/test/Piero_Baldini##mobyDick01_3sec.wav')
#v = Voiceid(db, '/Users/labcontenuti/Desktop/voiceid/DataSetVideolinaVoci/Giacomo_Mameli/2/Giacomo_Mameli##1-1.wav')
v = Voiceid(db, '/Users/labcontenuti/Documents/workspace/activevoice/audio_test/2sec.wav')
v.extract_speakers()
for c in v.get_clusters():
#print "c ", c
cluster = v.get_cluster(c)
"""
print "cluster.get_name ", cluster.get_name()
print "cluster.get_best_speaker ", cluster.get_best_speaker()
print "cluster.get_duration ", cluster.get_duration()
print "cluster.to_dict ", cluster.to_dict()
print "cluster.get_segments ", cluster.get_segments()
cluster.print_segments()
#cluster.print_segments()
"""
import sys
import speech_recognition as sr
fileStorage = sys.argv[1]
subprocess.check_output(["python", "speak.py", fileStorage]);
# create voice db
db = GMMVoiceDB('voiceDB')
print "DB Models:"
print db.get_speakers()
print
v = Voiceid(db, sys.argv[1])
v.extract_speakers(False, True, True)
# print "\nSpeakers:\n"
speakerName = "unknown"
for c in v.get_clusters():
cluster = v.get_cluster(c)
distance = cluster.get_distance()
speakerPrinted = False
# print distance
# print cluster.get_best_five()
# print
if (cluster.get_best_speaker() != "unknown" or abs(cluster.get_best_five()[0][1]) > 34):
# print cluster
from voiceid.sr import Voiceid
from voiceid.db import GMMVoiceDB
import sys
# create voice db
db = GMMVoiceDB('voiceDB')
print "DB Models:"
print db.get_speakers()
v = Voiceid(db, sys.argv[1])
v.extract_speakers(False, True)
# Set cluster speaker
c = v.get_cluster('S1')
c.set_speaker('derek')
c = v.get_cluster('S6')
c.set_speaker('derek')
# update db
v.update_db()
print "DB Models:"
print db.get_speakers()
def test_dir():
all_wav = []
for f in os.listdir(DIR_PATH):
if fnmatch.fnmatch(f, "*.wav") and f.find("##") > -1:
all_wav.append(f)
all_wav = verify_goodness(all_wav)
print all_wav
print "verifica ok... procediamo"
logging.basicConfig(filename=DIR_PATH + "/result.log", level=logging.INFO)
db = GMMVoiceDB("/Users/labcontenuti/.voiceid/gmm_db")
# filter_criteria="Fabrizio_Gifuni##02Pasticciaccio.wav"
for wav in all_wav:
# if wav.find(filter_criteria)>-1:
print "--------------------" + wav + "------------------------"
v = Voiceid(db, DIR_PATH + wav)
v.extract_speakers()
logging.info("\n\n\n---------------------" + wav + "------------------")
wav_tostring = str(wav) + "; "
for c in v.get_clusters():
cluster = v.get_cluster(c)
for seg in cluster.get_segments():
print "\n seg: ", seg
print seg.get_start()
print seg.get_end()
print humanize_time(seg.get_duration())
print "cluster.get_duration ", cluster.get_duration()
logging.info("duration: " + str(cluster.get_duration()))
wav_tostring = wav_tostring + ": " + "duration:" + str(cluster.get_duration())
print "cluster.get_name ", cluster.get_name()
print "cluster.get_best_speaker ", cluster.get_best_speaker(), " correct?", verify_speaker(
wav, cluster.get_best_speaker()
)
logging.info("best_speaker " + str(cluster.get_best_speaker()))
wav_tostring = wav_tostring + "; " + "best_speaker:" + str(cluster.get_best_speaker())
print "v._get_time() ", v._get_time(), " ", humanize_time(v._get_time())
logging.info("get_time: " + str(humanize_time(v._get_time())))
wav_tostring = wav_tostring + "; " + "get_time:" + str(humanize_time(v._get_time()))
# if cluster.get_best_speaker()=="unknown":
print "cluster.get_best_five ", cluster.get_best_five(), cluster.get_best_five()[0]
logging.info("best of five: " + str(cluster.get_best_five()[0]))
wav_tostring = wav_tostring + "; " + "best of five:" + str(cluster.get_best_five()[0])
print "cluster.get_distance() ", str(cluster.get_distance())
logging.info("distance best-closest " + str(cluster.get_distance()))
wav_tostring = wav_tostring + "; " + "distance best-closest:" + str(cluster.get_distance())
print
if not cluster.get_best_speaker() == "unknown":
if verify_speaker(wav, cluster.get_best_speaker()):
find_ok.append(wav_tostring)
else:
find_errata.append(wav_tostring)
else:
not_find.append(wav_tostring)
if verify_speaker(wav, cluster.get_best_five()[0][0]):
unknown_ok.append(wav_tostring)
else:
unknown_errata.append(wav_tostring)
print "Number TEST--" + str(len(find_ok) + len(find_errata) + len(not_find))
print "OK--------- " + str(len(find_ok))
print "ERRATA-------" + str(len(find_errata))
print "UNKNOWN------" + str(len(not_find))
print "UNKNOWN ERR--" + str(len(unknown_errata))
print "UNKNOWN OK--" + str(len(unknown_ok))
# resutl_file=file(DIR_PATH+"result.txt", "a+")
logging.info("\n\n") # for wav in unknown_errata:
logging.info("Number TEST " + str(len(find_ok) + len(find_errata) + len(not_find)))
logging.info("ok " + str(len(find_ok)))
logging.info("ERRATA" + str(len(find_errata)))
logging.info("UNKNOWN" + str(len(not_find)))
logging.info("UNKNOWN ERR " + str(len(unknown_errata)))
logging.info("UNKNOWN OK " + str(len(unknown_ok)))
logging.info("\n\n\n----UNKNOWN OK ---")
for f in unknown_ok:
logging.info(f)
logging.info("\n\n\n----UNKNOWN ERRATA ---")
for f in unknown_errata:
logging.info(f)
logging.info("\n\n\n----ERRATA---")
for f in find_errata:
logging.info(f)
if DIR_PATH:
for f in os.listdir(DIR_PATH):
if fnmatch.fnmatch(f, '*.wav') and f.find("##")>-1:
all_wav.append(DIR_PATH+"/"+f)
if FILE_PATH:
if fnmatch.fnmatch(FILE_PATH, '*.wav') and FILE_PATH.find("##")>-1:
all_wav.append(FILE_PATH)
print "all_wav ", all_wav
verify_goodness(all_wav)
print "verifica ok... procediamo"
db = GMMVoiceDB(DB_PATH)
for wav in all_wav:
OK=0
OK_best5=0
print "--------------------"+ wav+ "------------------------"
v = Voiceid(db, wav)
v.extract_speakers()
for c in v.get_clusters():
cluster = v.get_cluster(c)
print "cluster.get_best_speaker ", str(cluster.get_best_five()[0]) , " correct?", verify_speaker(wav, cluster.get_best_speaker())
if verify_identity(PID, str(cluster.get_best_five()[0])):
OK=OK+1
if verify_speaker(wav, cluster.get_best_five()[0] ):
OK_best5=OK_best5+1
if 2*OK>= len(v.get_clusters()):
print "------------------VERIFICA RIUSCITA------------------Doveva riuscire? ", 2*OK_best5>= len(v.get_clusters())
else:
print "+++++++++++++++++VERIFICA FALLITA+++++++++++++++++++++Doveva riuscire? ", 2*OK_best5>= len(v.get_clusters())
print "RND"
