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 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(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 train(evaluation):
n = len(evaluation['training'])
for i, example in enumerate(evaluation['training']):
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']):
'Error during training for {0}'.format(example_name)
continue
for word, speech in zip(example['words'], speech_words):
wav_name = id_generator() + '.wav'
save_wav(word, wav_name, rate, speech, '../dictionary')
print('Completed {0}%'.format(100.0 * (i + 1) / float(n)))
def train(evaluation):
n = len(evaluation['training'])
for i, example in enumerate(evaluation['training']):
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']):
'Error during training for {0}'.format(example_name)
continue
for word, speech in zip(example['words'], speech_words):
wav_name = id_generator() + '.wav'
save_wav(word, wav_name, rate, speech, '../dictionary')
print('Completed {0}%'.format(100.0 * (i + 1) / float(n)))
def train(dirs):
print('Training on {0}'.format(dirs))
reset_dir(DICTIONARY)
for i, dir in enumerate(dirs):
for file in os.walk(DIR_NAME + '/' + dir).next()[2]:
word = file.split('_')[0]
if not filter(word):
continue
rate, samples = wavfile.read('{0}/{1}/{2}'.format(
DIR_NAME, dir, file))
_, speech_words = extract_words(rate, samples, 1)
if len(speech_words) != 1:
print('Error extracting {0}/{1}'.format(dir, file))
continue
wav_name = id_generator() + '.wav'
save_wav(word, wav_name, rate, speech_words[0], DICTIONARY)
print('Completed {0}%'.format(100.0 * (i + 1) / float(len(dirs))))
print('Done training.')
def train(dirs):
print('Training on {0}'.format(dirs))
reset_dir(DICTIONARY)
for i, dir in enumerate(dirs):
for file in os.walk(DIR_NAME + '/' + dir).next()[2]:
word = file.split('_')[0]
if not filter(word):
continue
rate, samples = wavfile.read('{0}/{1}/{2}'.format(
DIR_NAME, dir, file))
_, speech_words = extract_words(rate, samples, 1)
if len(speech_words) != 1:
print('Error extracting {0}/{1}'.format(dir, file))
continue
wav_name = id_generator() + '.wav'
save_wav(word, wav_name, rate, speech_words[0], DICTIONARY)
print('Completed {0}%'.format(100.0 * (i + 1) / float(len(dirs))))
print('Done training.')
def test_word_extractor(evaluation):
n = len(evaluation['examples20'])
all = 0
correct = 0
for i, test in enumerate(evaluation['examples20']):
test_name = test['file']
hint = test['expected_words_count']
rate, samples = wavfile.read('../wav/{0}.wav'.format(test_name))
_, speech_words = extract_words(rate, samples, hint)
#_, speech_words_no_hint = extract_words(rate, samples)
speech_words_no_hint = []
print('Test {0}: expected = {1}, extracted = {2}, with no hint = {3}'.
format(test_name, test['expected_words_count'],
len(speech_words), len(speech_words_no_hint)))
all += 1
if len(speech_words) == test['expected_words_count']:
correct += 1
print('Completed {0}%'.format(100.0 * (i + 1) / float(n)))
print('Extracted {0}% correctly.'.format(100.0 * float(correct) /
float(all)))
def test_word_extractor(evaluation):
n = len(evaluation['examples20'])
all = 0
correct = 0
for i, test in enumerate(evaluation['examples20']):
test_name = test['file']
hint = test['expected_words_count']
rate, samples = wavfile.read('../wav/{0}.wav'.format(test_name))
_, speech_words = extract_words(rate, samples, hint)
#_, speech_words_no_hint = extract_words(rate, samples)
speech_words_no_hint = []
print('Test {0}: expected = {1}, extracted = {2}, with no hint = {3}'
.format(test_name,
test['expected_words_count'],
len(speech_words),
len(speech_words_no_hint)))
all += 1
if len(speech_words) == test['expected_words_count']:
correct += 1
print('Completed {0}%'.format(100.0 * (i + 1) / float(n)))
print('Extracted {0}% correctly.'.format(
100.0 * float(correct) / float(all)))
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)))