def record_and_test(textbox, button, filename="test_files/test.wav"):
# Buton on
button.configure(state="disabled")
textbox.configure(state="normal")
textbox.delete("1.0", END)
textbox.insert(INSERT, "Recording")
textbox.tag_add("recording", "1.0", END)
textbox.configure(state="disabled")
# Record to test
record_to_file(filename)
# Feed into ANN
testNet = testInit()
inputArray = extractFeature(filename)
print(len(inputArray))
outStr = feedToNetwork(inputArray, testNet)
# Button of
textbox.configure(state="normal")
textbox.delete("1.0", END)
textbox.tag_remove("recording", "1.0")
textbox.insert(INSERT, outStr)
textbox.tag_add("success", "1.0", END)
textbox.configure(state="disabled")
button.configure(state="normal")
def record_and_test(textbox, button, filename="test_files/test.wav"):
# Disable button and change text
button.configure(state="disabled")
textbox.configure(state="normal")
textbox.delete("1.0", END)
textbox.insert(INSERT, "Recording")
textbox.tag_add("recording", "1.0", END)
textbox.configure(state="disabled")
# Record to file
record_to_file(filename)
# Feed into ANN
testNet = testInit()
inputArray = extractFeature(filename)
print len(inputArray)
outStr = feedToNetwork(inputArray,testNet)
# Change text and re-enable button
textbox.configure(state="normal")
textbox.delete("1.0", END)
textbox.tag_remove("recording", "1.0")
textbox.insert(INSERT, outStr)
textbox.tag_add("success", "1.0", END)
textbox.configure(state="disabled")
button.configure(state="normal")
def record_and_train(textbox, button):
# button on
button.configure(state="disabled")
textbox.configure(state="normal")
textbox.delete("1.0", END)
textbox.insert(INSERT, "Recording")
textbox.tag_add("recording", "1.0", END)
textbox.configure(state="disabled")
words = ['down', 'eat', 'sleep', 'up']
for i in range(len(words)): # WORD loop
for j in range(10): # REPEAT loop
# record to train
record_to_file("training_sets/" + words[i] + "-" + str(j + 1) +
".wav")
print("repeat", words[i])
if (len(words) != i + 1):
print("next word", words[i + 1])
else:
print("finish")
# Button of
textbox.configure(state="normal")
textbox.delete("1.0", END)
textbox.tag_remove("recording", "1.0")
textbox.tag_add("success", "1.0", END)
textbox.configure(state="disabled")
button.configure(state="normal")
# MFCC calculation and TRAIN
mfcc_apply()
print("train start..")
TRAIN()
def test_with_recording():
import wave
from record import record_to_file, play_sound_file
test_audio_file = f"{current_dir}/tmp/api_test.wav"
if not os.path.isfile(test_audio_file):
record_to_file(test_audio_file, sampling_freq=10000)
play_sound_file(test_audio_file)
result = send_to_speech_api(test_audio_file)
print(result)
def main():
import wave
from record import record_to_file, play_sound_file
test_audio_file = f"{current_dir}/tmp/api_test.wav"
if not os.path.isfile(test_audio_file):
record_to_file(test_audio_file, sampling_freq=16000)
play_sound_file(test_audio_file)
result = transcribe_file(f"{current_dir}/tmp/api_test.wav")
print(result)
def record_some_words():
sleep(2)
speak("Bitte sprechen Sie nach dem Piepston")
beep = Beep()
beep.play(1000)
record_to_file('demo.wav')
speak("Danke! Sie haben folgende Nachricht aufgenommen:")
play_from_file('demo.wav')
speak("... und noch einmal:")
play_from_file('demo.wav')
def test_with_live_recording(duration=1, sampling_rate=10000):
"""
Test the API by recording a digit using the microphone and then preprocessing it, sending the spectrogram and awaiting the result.
"""
from time import sleep
from make_spectrograms import make_spectrogram_from_wav_file
from record import record_to_file, play_sound_file, record, write_to_file
test_audio_file = f"{current_dir}/tmp/api_test.wav"
test_img_path = f"{current_dir}/tmp/api_test.png"
# do a quick countdown before recording.
for i in reversed(range(4)):
print(f"\rRecording in {i}", end="\r")
sleep(1)
record_to_file(test_audio_file,
recording_length_secs=duration,
sampling_freq=sampling_rate)
play_sound_file(test_audio_file)
result = send_to_speech_api(test_audio_file)
print(result)
def record_and_test(textbox, button, filename):
button.configure(state="disabled")
textbox.configure(state="normal")
textbox.delete("1.0", END)
textbox.insert(INSERT, "Recording")
textbox.tag_add("recording", "1.0", END)
textbox.configure(state="disabled")
record_to_file(filename)
testNet = testInit()
inputArray = extractFeature(filename)
print len(inputArray)
outStr = "Detected : " + str(feedToNetwork(inputArray, testNet) + 1)
textbox.configure(state="normal")
textbox.delete("1.0", END)
textbox.tag_remove("recording", "1.0")
textbox.insert(INSERT, outStr)
textbox.tag_add("success", "1.0", END)
textbox.configure(state="disabled")
button.configure(state="normal")
def record_and_test(textbox, button, filename):
button.configure(state="disabled")
textbox.configure(state="normal")
textbox.delete("1.0", END)
textbox.insert(INSERT, "Recording")
textbox.tag_add("recording", "1.0", END)
textbox.configure(state="disabled")
record_to_file(filename)
testNet = testInit()
inputArray = extractFeature(filename)
print len(inputArray)
outStr = "Detected : " + str(feedToNetwork(inputArray,testNet)+1)
textbox.configure(state="normal")
textbox.delete("1.0", END)
textbox.tag_remove("recording", "1.0")
textbox.insert(INSERT, outStr)
textbox.tag_add("success", "1.0", END)
textbox.configure(state="disabled")
button.configure(state="normal")
def recordSoundData(self):
record.record_to_file("/tmp/%s.lyric" % self.hash)
import playsound as plsnd
import requests as req
HOST = "http://192.168.0.103:3000"
if __name__ == '__main__':
testNet = testInit()
num_loop = 0
filename = "test_files/test.wav"
while True:
# Record to file
num_loop += 1
print("please speak a word into the microphone", num_loop)
record_to_file(filename)
# Feed into ANN
inputArray = extractFeature(filename)
res = feedToNetwork(inputArray, testNet)
if (res == 0):
# ban can giup gi?
plsnd.playsound("speak_out_files/bancangiupgi.wav")
print("Ban can giup gi? ...")
record_to_file(filename)
inputArray = extractFeature(filename)
res = feedToNetwork(inputArray, testNet)
def run_record(status):
record_to_file('query.wav')
status.audio = True
status.stop_recording()
## здесь соберем все воедино
import acrcloud_api
import vk_api
import record
# авторизовываемся в вк
api = vk_api.authorization(access_token=vk_api.access_token)
# получаем путь к файлу от алексы...
#но пока стандартный
music_file_path = '/Users//Desktop/Git/icon_hack/demo.mp3' # запись в файл wave.mp3
record.record_to_file(music_file_path) # запись с микрофона
# обрабатываем его через acrcloud_api
responce = acrcloud_api.get_responce(config=acrcloud_api.config,
music_file_path=music_file_path,
start_seconds=3)
title, artist = acrcloud_api.parce_responce(responce)
# отправляем шаблонную строку в вк
pattern_string = "Это песня " + title + " исполнителя " + artist + "!"
# еще хз как выбирать юзера, поэтому отправляю себе
#vk_api.send_message(api=api, user_id=62811131, message=pattern_string)
def RecordSpeech(self):
print "please speak a word into the microphone"
record_to_file(join(dirname(__file__),self.audiofile))
print "done - result written to audio.wav"
import sys import record if __name__ == '__main__': record.record_to_file(sys.stdout)
def exp_evidence(logits):
return np.exp(np.clip(logits, -10, 10))
def uncertainty_score(logits_output):
evidence = exp_evidence(logits_output)
alpha = evidence + 1
u_score = 10 / np.sum(alpha, axis=1, keepdims=True) # K = num_classes = 10
return u_score
while True:
print("Listening...")
record_start = time()
record_to_file("sound/input.wav")
record_time = time() - record_start
work_dir = 'sound'
start_time = time()
test_x, test_y = data_processing(work_dir)
load_time = time() - start_time
# # Testing on loaded model
start_time = time()
output = sess.run(output_tensor, feed_dict={model_input: test_x})
y_pred = predict_label(output)
uncertainty_y_list = uncertainty_score(output)
test_time = time() - start_time
print("Inference time:", infer_time)
