def initialise():
# Initialise the speech recognition engine and the iCub controller
my_speech = SpeechRecognizer(
hmm_path=
"/home/massimiliano/pyERA/examples/ex_icub_trust_cognitive_architecture/sphinx/model/en-us/en-us",
language_model_path=
"/home/massimiliano/pyERA/examples/ex_icub_trust_cognitive_architecture/sphinx/model/en-us/en-us.lm.bin",
dictionary_path=
"/home/massimiliano/pyERA/examples/ex_icub_trust_cognitive_architecture/sphinx/data/icub.dic",
grammar_path=
"/home/massimiliano/pyERA/examples/ex_icub_trust_cognitive_architecture/sphinx/data/icub.gram",
rule_name='icub.basicCmd',
fsg_name="icub")
# iCub initialization
my_icub = iCub(icub_root='/icubSim')
# Load acapela configuration from file
my_icub.set_acapela_credential("./acapela_config.csv")
account_login, application_login, application_password, service_url = my_icub.get_acapela_credential(
)
print("[ACAPELA]Acapela configuration parameters:")
print("Account Login: "******"Application Login: "******"Account Password: "******"Service URL: " + str(service_url))
print("")
# Return the objects
return my_speech, my_icub
def test(dirs):
print('Testing on {0}'.format(dirs))
print('Loading recognizer, please wait...')
speech_recognizer = SpeechRecognizer(DICTIONARY)
print('Loading completed.')
all = 0
correct = 0
for i, dir in enumerate(dirs):
curr_all = 0
curr_correct = 0
for file in os.walk(DIR_NAME + '/' + dir).next()[2]:
expected_word = file.split('_')[0]
if not filter(expected_word):
continue
rate, samples = wavfile.read('{0}/{1}/{2}'.format(
DIR_NAME, dir, file))
rate, speech_words = extract_words(rate, samples, 1)
if len(speech_words) != 1:
print('Error extracting {0}/{1}'.format(dir, file))
continue
word, _ = try_recognition(rate, speech_words[0], speech_recognizer)
all += 1
curr_all += 1
if word == expected_word:
correct += 1
curr_correct += 1
else:
print 'Expected {0}, got {1}'.format(expected_word, word)
if curr_all > 0:
print('Correct {0}% in {1}'.format(100.0 * curr_correct / curr_all,
dir))
print('Completed {0}%'.format(100.0 * (i + 1) / float(len(dirs))))
print('--------------------------------')
print('Correct {0}%'.format(100.0 * correct / all))
print('Done testing.')
def test(evaluation):
print('Loading recognizer, please wait...')
speech_recognizer = SpeechRecognizer('../dictionary/')
print('Loading completed.')
n = len(evaluation['test'])
all = 0
correct = 0
for i, example in enumerate(evaluation['test']):
curr_all = 0
curr_correct = 0
example_name = example['file']
hint = example['expected_words_count']
rate, samples = wavfile.read('../wav/{0}.wav'.format(example_name))
rate, speech_words = extract_words(rate, samples, hint)
if len(speech_words) != len(example['words']):
print('Error in extracting words. Skipping {0}.'.format(
example_name))
continue
for word, sol_word in zip(speech_words, example['words']):
sol, _ = try_recognition(rate, word, speech_recognizer)
#print('Expected {0}, recognized {1}'.format(sol_word, sol))
all += 1
curr_all += 1
if sol == sol_word:
correct += 1
curr_correct += 1
print('Completed {0}%'.format(100.0 * (i + 1) / float(n)))
print('Stats on {0}, correct {1}%'.format(
example_name, 100.0 * curr_correct / curr_all))
break
print('--------------------------------------')
print('Correct {0}%'.format(100.0 * correct / all))
def initialise(root='/icubSim'):
# Initialise the speech recognition engine and the iCub controller
my_speech = SpeechRecognizer(hmm_path=HMM_PATH,
language_model_path=LANGUAGE_MODEL_PATH,
dictionary_path=DICTIONARY_PATH,
grammar_path=GRAMMAR_PATH,
rule_name='icub.basicCmd',
fsg_name="icub")
# iCub initialization
my_icub = iCub(icub_root=root)
# Load acapela configuration from file
my_icub.set_acapela_credential(ACAPELA_CSV)
account_login, application_login, application_password, service_url = my_icub.get_acapela_credential(
)
print("[ACAPELA]Acapela configuration parameters:")
print("Account Login: "******"Application Login: "******"Account Password: "******"Service URL: " + str(service_url))
print("")
# Return the objects
return my_speech, my_icub
def initialise(objects_path):
# Initialise the speech recognition engine and the iCub controller
my_speech = SpeechRecognizer(
hmm_path=
"/home/massimiliano/pyERA/examples/ex_icub_trust_cognitive_architecture/sphinx/model/en-us/en-us",
language_model_path=
"/home/massimiliano/pyERA/examples/ex_icub_trust_cognitive_architecture/sphinx/model/en-us/en-us.lm.bin",
dictionary_path=
"/home/massimiliano/pyERA/examples/ex_icub_trust_cognitive_architecture/sphinx/data/icub.dic",
grammar_path=
"/home/massimiliano/pyERA/examples/ex_icub_trust_cognitive_architecture/sphinx/data/icub.gram",
rule_name='icub.basicCmd',
fsg_name="icub")
# iCub initialization
my_icub = iCub(icub_root='/icubSim')
# Load acapela configuration from file
my_icub.set_acapela_credential("./acapela_config.csv")
account_login, application_login, application_password, service_url = my_icub.get_acapela_credential(
)
print("[ACAPELA]Acapela configuration parameters:")
print("Account Login: "******"Application Login: "******"Account Password: "******"Service URL: " + str(service_url))
print("")
#If there are images of template stored in the folder,
#it opens the folder and load the objects in memory.
print("[STATE init] loading objects from memory...")
for file in os.listdir(objects_path):
if file.endswith(".png"):
object_name = os.path.splitext(file)[0]
print("Object name: " + object_name)
print("Object path: " + os.path.join(objects_path, file))
template = cv2.imread(os.path.join(objects_path, file))
my_icub.learn_object_from_histogram(template, object_name)
# Return the objects
return my_speech, my_icub
import sys
from speech_recognition import SpeechRecognizer
if __name__ == '__main__':
try:
recog = SpeechRecognizer()
except SpeechRecognizer.InvalidDevice as e:
print(e.message)
sys.exit(1)
def main():
warnings.simplefilter('ignore')
clear_tmp()
print('0. Simplified')
print('1. Full')
option = input('Select input: ')
if option == 0:
dir = 'simplified_dict/'
else:
dir = 'dictionary/'
print('Loading recognizer, please wait...')
speech_recognizer = SpeechRecognizer(dir)
print('Loading completed.')
print('0. Record new')
print('1. Use existing wav file (from wav/)')
print('2. Just record wav for later use')
print('3. Split existing file (from wav/)')
option = input('Select input: ')
if option == 0:
speech_file = record_wav()
elif option == 1:
speech_file = raw_input('File name (without .wav): ')
elif option == 2:
_ = record_wav()
return
elif option == 3:
speech_file = raw_input('File name (without .wav): ')
hint = input(
'Hint maybe (enter expected number of words, 0 for no hint)? ')
rate, samples = wavfile.read('wav/{0}.wav'.format(speech_file))
rate, speech_words = extract_words(rate, samples, hint)
for speech in speech_words:
_ = save_tmp_wav(rate, speech)
print('Extracted {0} words'.format(len(speech_words)))
return
else:
print('Invalid option')
return
hint = input('Hint maybe (enter expected number of words, 0 for no hint)? ')
rate, samples = wavfile.read('wav/{0}.wav'.format(speech_file))
rate, speech_words = extract_words(rate, samples, hint)
print('Extracted {0}'.format(len(speech_words)))
correct_count = 0
count = 0
for speech in speech_words:
speech_copy = speech[:]
_ = raw_input('Press enter to continue...')
wav_name = save_tmp_wav(rate, speech)
play_wav('tmp/{0}'.format(wav_name))
word, confidence = try_recognition(rate, speech_copy, speech_recognizer)
print('Recognized [{0}] with confidence {1}'.format(word, confidence))
count += 1
correct = input('Is it correct (0/1)? ')
if correct == 0:
add = input('Do you want to add this word in dictionary (0/1)? ')
if add == 0:
print(':(')
else:
text_word = raw_input('Write string representation: ')
text_word.rstrip('\n').replace(' ', '_')
save_wav(text_word, wav_name, rate, speech)
print('Added new word [{0}]'.format(text_word))
else:
correct_count += 1
os.rename('tmp/{0}'.format(wav_name), 'tmp/{0}.wav'.format(word))
print(':)')
print('Recognized {0}% correctly.'.format(
100.0 * float(correct_count) / float(count)))