def test_basic_slm():
asr = SpeechRecognizer(url, **asr_kwargs)
asr.recognize(FileAudioSource(phone_wav), LanguageModelList(slm))
result = asr.wait_recognition_result()[0]
asr.close()
assert len(result.alternatives[0]['text']) > 0
assert int(result.alternatives[0]['score']) > 90
def test_cancel_on_recognize():
asr = SpeechRecognizer(url, **asr_kwargs)
asr.recognize(DelayedFileAudioSource(phone_wav),
LanguageModelList(phone_grammar_uri))
time.sleep(2)
asr.cancel_recognition()
result = asr.wait_recognition_result()
assert len(result) == 0
def test_no_match():
asr = SpeechRecognizer(url, **asr_kwargs)
asr.recognize(FileAudioSource(silence_wav), LanguageModelList(slm))
result = asr.wait_recognition_result()[0]
asr.close()
# NO_SPEECH occurs with enabled endpointer, and NO_MATCH with disabled
assert result.result_code in ("NO_SPEECH", "NO_MATCH"), \
"Result code is {}".format(result.resultCode)
def test_recognize_buffer_audio_source():
asr = SpeechRecognizer(url, **asr_kwargs)
asr.recognize(DelayedFileAudioSource(phone_wav),
LanguageModelList(phone_grammar_uri))
result = asr.wait_recognition_result()[0].alternatives[0]
asr.close()
assert len(result['text']) > 0
assert len(result['interpretations']) > 0
assert int(result['score']) > 90
def test_wait_recognition_result_duplicate():
asr = SpeechRecognizer(url, **asr_kwargs)
asr.recognize(FileAudioSource(phone_wav),
LanguageModelList(phone_grammar_uri))
result = asr.wait_recognition_result()[0].alternatives[0]
result_empty = asr.wait_recognition_result()
asr.close()
assert len(result['text']) > 0
assert len(result['interpretations']) > 0
assert int(result['score']) > 90
assert len(result_empty) == 0
def test_no_input_timeout():
config = {
'decoder.startInputTimers': 'true',
'noInputTimeout.value': '100',
'noInputTimeout.enabled': 'true',
}
asr = SpeechRecognizer(url, recog_config=config, **asr_kwargs)
asr.recognize(DelayedFileAudioSource(silence_wav),
LanguageModelList(phone_grammar_uri))
result = asr.wait_recognition_result()
asr.close()
assert(result[0].result_code in ("NO_INPUT_TIMEOUT", "NO_MATCH"))
def test_multiple_recognize():
asr = SpeechRecognizer(url, **asr_kwargs)
results = []
for i in range(3):
asr.recognize(DelayedFileAudioSource(phone_wav),
LanguageModelList(phone_grammar_uri))
results.append(asr.wait_recognition_result()[0].alternatives[0])
asr.close()
for result in results:
assert len(result['text']) > 0
assert len(result['interpretations']) > 0
assert int(result['score']) > 90
def test_max_wait_seconds_thread_response():
asr = SpeechRecognizer(url, max_wait_seconds=2, **asr_kwargs)
asr.recognize(DelayedFileAudioSource(phone_wav),
LanguageModelList(phone_grammar_uri))
try:
asr.wait_recognition_result()
asr.close()
except RecognitionException as e:
asr.close()
assert e.code == "FAILURE"
else:
asr.close()
assert False
def test_basic_grammar():
asr = SpeechRecognizer(url, **asr_kwargs)
asr.recognize(FileAudioSource(phone_wav),
LanguageModelList(phone_grammar_uri))
result = asr.wait_recognition_result()
asr.close()
alt = None
for res in result:
if len(res.alternatives) > 0:
alt = res.alternatives[0]
break
assert alt is not None
assert len(alt['text']) > 0
assert len(alt['interpretations']) > 0
assert int(alt['score']) > 90
def test_inline_grammar():
with open(yes_grammar_path) as f:
body = f.read()
asr = SpeechRecognizer(url, **asr_kwargs)
asr.recognize(FileAudioSource(yes_wav),
LanguageModelList(("yes_no", body)))
result = asr.wait_recognition_result()
asr.close()
alt = None
for res in result:
if len(res.alternatives) > 0:
alt = res.alternatives[0]
break
assert alt is not None
assert len(alt['text']) > 0
assert len(alt['interpretations']) > 0
assert int(alt['score']) > 90
def test_duplicate_recognize():
asr = SpeechRecognizer(url, **asr_kwargs)
asr.recognize(DelayedFileAudioSource(phone_wav),
LanguageModelList(phone_grammar_uri))
time.sleep(2)
try:
asr.recognize(DelayedFileAudioSource(phone_wav),
LanguageModelList(phone_grammar_uri))
except RecognitionException as e:
assert e.code == "FAILURE"
else:
assert False
result = asr.wait_recognition_result()[0].alternatives[0]
asr.close()
assert len(result['text']) > 0
assert len(result['interpretations']) > 0
assert int(result['score']) > 90
def recognize_worker(url, asr_kwargs, wav_path, lm_list, recognitions,
executions, assertion):
error_msg = ("Parameters: url:{}, wav_path:{}, lm_list:{}".format(
url, wav_path, lm_list))
error_msg += "\nOn execution {}/{}\nOn recognition {}/{}"
error_msg += "\nError from assertion: {}"
for e in range(executions):
beg = time()
asr = SpeechRecognizer(url, **asr_kwargs)
for r in range(recognitions):
asr.recognize(FileAudioSource(wav_path), lm_list)
success, msg = assertion(asr.wait_recognition_result())
if not success:
error_msg = error_msg.format(executions, e, recognitions, r,
msg)
assert success, error_msg
asr.close()
asr._logger.info("[TIMER] TotalTime: {} s".format(time() - beg))
def test_equivalence_file_buffer():
# File
asr = SpeechRecognizer(url, **asr_kwargs)
asr.recognize(FileAudioSource(phone_wav),
LanguageModelList(slm))
result_file = asr.wait_recognition_result()[0].alternatives[0]['text']
asr.close()
# Buffer
asr = SpeechRecognizer(url, **asr_kwargs)
source = BufferAudioSource()
asr.recognize(source,
LanguageModelList(slm))
sig, rate = sf.read(phone_wav)
source.write((sig * 2**15).astype('int16').tobytes())
source.finish()
result_buffer = asr.wait_recognition_result()[0].alternatives[0]['text']
asr.close()
assert result_file == result_buffer
if __name__ == "__main__":
ostream = open('log.txt', 'a')
argc = len(argv)
if argc != 4 and argc != 6:
usage()
url = argv[1]
if os.path.isfile(argv[2]):
lm = LanguageModelList(
LanguageModelList.grammar_from_path('asdasdas', argv[2]))
else:
lm = LanguageModelList(LanguageModelList.from_uri(argv[2]))
apath = argv[3]
credentials = ("", "")
if argc == 6:
credentials = (argv[4], argv[5])
asr = SpeechRecognizer(url,
credentials=credentials,
log_stream=ostream,
log_level="debug",
max_wait_seconds=600)
asr.recognize(FileAudioSource(apath), lm)
res = asr.wait_recognition_result()
if res:
for k in res:
print(k.alternatives)
else:
print("Empty result! Check log.txt for more info.")
asr.close()
else:
lm = LanguageModelList(
LanguageModelList.from_uri(argv[2])
)
apath = argv[3]
credentials = ("", "")
if argc == 6:
credentials = (argv[4], argv[5])
asr = SpeechRecognizer(url, credentials=credentials,
log_stream=ostream,
log_level="debug",
max_wait_seconds=600)
source = BufferAudioSource()
asr.recognize(source, lm)
audio, rate = sf.read(apath)
# From float32 to int16, and then to raw bytes
source.write((audio * 2**15).astype('int16').tobytes())
source.finish()
res = asr.wait_recognition_result()
if res:
for k in res:
print(k.alternatives)
else:
print("Empty result! Check log.txt for more info.")
asr.close()
LanguageModelList.grammar_from_path(os.path.basename(argv[2]),
argv[2]))
else:
lm = LanguageModelList(LanguageModelList.from_uri(argv[2]))
credentials = ("", "")
if argc == 5:
credentials = (argv[3], argv[4])
asr = SpeechRecognizer(url,
credentials=credentials,
log_stream=ostream,
listener=PrinterListener(),
log_level="warning")
with MicAudioSource() as mic:
try:
asr.recognize(mic, lm)
result = asr.wait_recognition_result()
if result:
if result[0].resultCode == "RECOGNIZED":
print(result[0].alternatives[0]['text'])
else:
print(result[0].resultCode)
except KeyboardInterrupt:
print("Caught interrupt. Closing ASR instance...", file=stderr)
try:
asr.cancel_recognition()
except RecognitionException:
pass # Ignores exceptions on canceling
asr.close()