def do_recognition(self, fname, actual_rate, start_time, end_time):
request_id = uuid.uuid4()
self.querystring['requestid'] = self.querystring['requestid'].format(request_id = request_id)
self.headers['content-type'] = self.headers['content-type'].format(actual_rate = actual_rate)
# Since token is updated automatically it should always be valid
self.headers['authorization'] = self.headers['authorization'].format(token = self.oauth.token)
response = requests.request("POST",
self.url, # "http://httpbin.org/post",
headers = self.headers,
params = self.querystring,
data = open(fname, 'rb')
)
os.unlink(fname)
# print(response.text)
confidence = 0.0
if response.status_code == 200 and response.json()['header']['status'] == 'success':
a = response.json()
if 'lexical' in a['results'][0]:
# text = a['results'][0]['lexical']
text = a['results'][0]['name']
rospy.loginfo("text: {}".format(text))
if 'confidence' in a['results'][0]:
confidence = float(a['results'][0]['confidence'])
confidence = confidence * 100
rospy.loginfo("confidence: {}".format(confidence))
msg = SpeechStamped()
msg.header.stamp = start_time
msg.header.frame_id = self.frame_id
msg.duration = end_time - start_time
msg.text = text
msg.confidence = confidence
self.pub_speech.publish(msg)
# lock
self.lock.acquire()
try:
# modify the list
self.threads_list.remove(threading.currentThread())
finally:
# release
self.lock.release()
def do_recognition(self):
"""Do speech recognition"""
while self.started:
sox_p = subprocess.call(self.sox_args)
end_time = rospy.Time.now()
audio_len, _dummy_err = subprocess.Popen(self.length_args, stdout=subprocess.PIPE).communicate()
start_time = end_time - rospy.Duration(float(audio_len.strip()))
if float(audio_len) < self.dur_threshold:
rospy.logwarn("Recorded audio is too short ({}s < {}s). Ignoring".format(float(audio_len), self.dur_threshold))
continue
actual_rate, _dummy_err = subprocess.Popen(self.rate_args, stdout=subprocess.PIPE).communicate()
self.actual_rate = int(actual_rate.strip())
self.wget_cmd = self.wget_cmd.format(actual_rate=self.actual_rate, api_key=self.api_key, lang=self.lang)
self.wget_args = shlex.split(self.wget_cmd)
wget_out, wget_err = subprocess.Popen(
self.wget_args, stdout=subprocess.PIPE, stderr=subprocess.PIPE
).communicate()
# print wget_out
# print wget_err
confidence = 0.0
if not wget_err and len(wget_out) > 16:
wget_out = wget_out.split('\n', 1)[1]
a = json.loads(wget_out)['result'][0]
if 'transcript' in a['alternative'][0]:
text = a['alternative'][0]['transcript']
rospy.loginfo("text: {}".format(text))
if 'confidence' in a['alternative'][0]:
confidence = a['alternative'][0]['confidence']
confidence = confidence * 100
rospy.loginfo("confidence: {}".format(confidence))
msg = SpeechStamped()
msg.header.stamp = start_time
msg.header.frame_id = "human_frame"
msg.duration = end_time - start_time
msg.text = text
msg.confidence = confidence
self.pub_speech.publish(msg)
