import os import time import threading import wave import pyaudio import noisereduce from scipy.io import wavfile import webrtcvad from config import cfg CHUNK = 480 # frame count in buffer. webrtcvad just support 10ms, 20ms, 30ms chunk, so use 30ms( 16000 * 30 / 1000 ) FORMAT = pyaudio.paInt16 # bitsize of sampling CHANNELS = 1 # voice channel RATE = 16000 # bitrate of sampling MAXRECORDTIME =10 # the max record time of one record action MINSPEECHCOUNT =10 # the min speech chunk count in one sentence MAXEMPTYCOUNT =60 # the max empty chunk count in one sentence RECORDTIME =3 # time duration of one record buffer _recordthreadsinglelock = False class _recordthread(threading.Thread): def __init__(self): self.running = False self.directory = "" self.prename = "" self.bufsize = 0 self.looptime = 0.0 self.recordtime = 0.0 threading.Thread.__init__(self) def run(self): global _recordthreadsinglelock if _recordthreadsinglelock: return _recordthreadsinglelock = True self.running = True bufcount = 1 while self.running : start_time = time.time() filename = "%s%d.wav" % (self.prename, bufcount) filepath = os.path.join(self.directory, filename) hr = Recorder.RecordAudios(filepath, self.recordtime) if hr < 0: break end_time = time.time() jobduration = end_time - start_time (quo, rem) = divmod(jobduration, self.looptime) bufcount = (bufcount + quo + 1) % self.bufsize if bufcount == 0 : bufcount = self.bufsize waittime = self.looptime - rem if waittime > 0 : time.sleep(waittime) self.running = False # for break _recordthreadsinglelock = False def IsRunning(self): return _recordthreadsinglelock def StopThread(self): self.running = False while(_recordthreadsinglelock): time.sleep(1) class __recorder(): def __init__(self): self.pa = None self.stream = None self.rthread = None self.isrecording = False self.micindex = cfg.getint("device","micindex") self.vad = webrtcvad.Vad(1) def __del__(self): if self.stream is not None: if self.rec.isOpened(): self.CloseCamera() def IsOpened(self): if self.stream is None: return False return self.stream.is_active() def IsRecording(self): if self.isrecording: return True if self.rthread is not None: if self.rthread.IsRunning(): return True return False def OpenRecorder(self): if self.IsOpened(): return True self.pa = pyaudio.PyAudio() self.stream = self.pa.open(format=FORMAT, channels=CHANNELS, rate=RATE, input=True, frames_per_buffer=CHUNK, input_device_index=self.micindex) if self.IsOpened(): return True else: print("Zhi Error: Cann't open mic") return False def CloseRecorder(self): if self.IsOpened(): if self.rthread is not None: self.rthread.StopThread() self.rthread = None self.stream.stop_stream() self.stream.close() self.pa.terminate() self.stream = None self.pa = None ''' humanonly: only record hunman vioce filepath: save record file path maxtime: the max record time of one sentence (second) ''' def RecordAudio(self, humanonly, filepath, maxtime=MAXRECORDTIME): if not self.IsOpened(): return -1 if self.IsRecording(): return -2 self.isrecording = True # read data into buffer, and check if it is speech and if this sentence is finish start_sentence = False continuous_count = 0 continuous_buf = [] empty_count = 0 record_buf = [] for _ in range(0, int(RATE / CHUNK * maxtime)): data = self.stream.read(CHUNK, exception_on_overflow = False) # if just record human voice, just "is_speach" data can be record. those not human vioce, will be watch as empty data. if humanonly : if not self.vad.is_speech(data, RATE): empty_count += 1 continuous_count = 0 if start_sentence: if empty_count > MAXEMPTYCOUNT : # if empty count in sentence is more than MAXEMPTYCOUNT, that means sentence finish break else: record_buf.append(data) continue # record data into record buffer continuous_count += 1 if continuous_count == MINSPEECHCOUNT: # if continous count equire to min speech chunk count, means data in continuous buffer available, move date into record buffer start_sentence = True record_buf.extend(continuous_buf) continuous_buf = [] empty_count = 0 elif continuous_count > MINSPEECHCOUNT: # if coutinous count is more then min speech chunk count, write data into record buffer directly record_buf.append(data) else: # if coutinous count is less then min speech chunk count, just write data into continuous buffer continuous_buf.append(data) empty_count += 1 # empty_count still add 1, because is still un-abailable if len(record_buf) > 0 : # write to file with wave.open(filepath, 'wb') as wf: wf.setnchannels(CHANNELS) wf.setsampwidth(self.pa.get_sample_size(FORMAT)) wf.setframerate(RATE) wf.writeframes(b''.join(record_buf)) # reduce noise, and rewrite to file _, data = wavfile.read(filepath) reduced_noise = noisereduce.reduce_noise(y=data, sr=RATE) wavfile.write(filepath, RATE, reduced_noise) self.isrecording = False if len(record_buf) == 0 : return -3 else : return 0 ''' filepath: save record file path time: the duration time if record (second) ''' def RecordAudios(self, filepath, time=RECORDTIME): if not self.IsOpened(): return -1 # read data into buffer record_buf = [] for _ in range(0, int(RATE / CHUNK * time)): data = self.stream.read(CHUNK, exception_on_overflow = False) record_buf.append(data) # write to file with wave.open(filepath, 'wb') as wf: wf.setnchannels(CHANNELS) wf.setsampwidth(self.pa.get_sample_size(FORMAT)) wf.setframerate(RATE) wf.writeframes(b''.join(record_buf)) # reduce noise, and rewrite to file _, data = wavfile.read(filepath) reduced_noise = noisereduce.reduce_noise(y=data, sr=RATE) wavfile.write(filepath, RATE, reduced_noise) return 0 def RecordAudiosStart(self, directory, prename, bufsize, looptime, recordtime): if not self.IsOpened(): return -1 if self.IsRecording(): return -2 if self.rthread is None : self.rthread=_recordthread() self.rthread.directory = directory self.rthread.prename = prename self.rthread.bufsize = bufsize self.rthread.looptime = looptime self.rthread.recordtime = recordtime self.rthread.start() return 0 else: print("Zhi Error: Record is bock when start it") return -3 def RecordAudiosStop(self): if self.rthread is None: print("Zhi Error: Record is not inited when stop it") return -1 if not self.rthread.IsRunning(): print("Zhi Error: Record is not started when stop it") return -2 self.rthread.StopThread() self.rthread = None return 0 Recorder = __recorder()