做啤酒行业的网站,设计外贸网站,国外知名设计网站大全,centos6安装wordpress本文最早发表于电子发烧友论坛#xff1a;【新提醒】【正点原子i.MX93开发板试用连载体验】基于深度学习的语音本地控制 - 正点原子学习小组 - 电子技术论坛 - 广受欢迎的专业电子论坛! (elecfans.com)
接下来就是要尝试训练中文提示词。首先要进行语料采集#xff0c;这是一… 本文最早发表于电子发烧友论坛【新提醒】【正点原子i.MX93开发板试用连载体验】基于深度学习的语音本地控制 - 正点原子学习小组 - 电子技术论坛 - 广受欢迎的专业电子论坛! (elecfans.com)
接下来就是要尝试训练中文提示词。首先要进行语料采集这是一个比较耗费人力的事情通常大公司会有有专人进行语料收集我只好自己亲自做。这里参考了AliOS Things里面提供的一个录音工具方便快速录音。对这个工具做了一点修改原来的代码只能在Linux下运行现在改成在Windows下也能运行。 import pyaudio
import wave
import random
import time
import os
from IPython import display
#from pydub import AudioSebment
#from pydub.playback import play
#from playsound import playsound
CHUNK 2
FORMAT pyaudio.paInt16
CHANNELS 1
RATE 16000
SAMPLEWIDTH 2
RECORD_SECONDS 1
FILE_FORMAT .wav
RECODER_NAME lk
#play stream
def play_wav(name, pyaudio):f wave.open(name,rb) #open stream play_stream pyaudio.open(format p.get_format_from_width(f.getsampwidth()), channels f.getnchannels(), rate f.getframerate(), output True) #read data data f.readframes(CHUNK)while data: play_stream.write(data)data f.readframes(CHUNK)#stop streamplay_stream.stop_stream() play_stream.close() #close PyAudio
# pyaudio.terminate()f.close()
def save_wav(name, frames):wf wave.open(name, wb)wf.setnchannels(CHANNELS)wf.setsampwidth(p.get_sample_size(FORMAT))wf.setframerate(RATE)wf.writeframes(b.join(frames))wf.close()
def record_wav(duration):time.sleep(0.2) # 1sec, 0.1secprint(开始录音,请说话......)
# count 3
# for i in range(3):
# time.sleep(0.2) # 1sec, 0.1sec
# count - 1
# print(count)frames []stream p.open(formatFORMAT,channelsCHANNELS,rateRATE,inputTrue,frames_per_bufferCHUNK)for i in range(0, int(RATE * duration / SAMPLEWIDTH)):data stream.read(CHUNK, exception_on_overflow False)frames.append(data)#count 0#while count int(RECORD_SECONDS * RATE):# data stream.read(CHUNK)# frames.append(data)# count CHUNKstream.stop_stream()stream.close()print(录音结束!)return frames
# main function
if __name__ __main__:p pyaudio.PyAudio()# input(请按回车键开始录制\n)# record filescount 0for i in range(250):input(请按回车键开始录制\n)print(开始第%d录制 % count)hash_name str(hex(abs(hash(RECODER_NAME str(random.random()))) % 1000000000)).replace(0x,) \ _nohash_ str(count) FILE_FORMATrframes record_wav(1) # record 1 secsave_wav(hash_name, rframes)#time.sleep(0.5) # 1sec, 0.1secprint(录音回放开始\n)play_wav(hash_name, p)print(录音回放结束\n)value input(按‘回车’保存放弃本条请按‘其他’键并回车!\n)if (value ):count 1print(保存录音成功)else:os.remove(hash_name)print(已删除本条录音)#display.display(display.Audio(hash_name, rate16000))#wav AudioSegment.from_wav(hash_name)#play(wav)p.terminate()
