python中文语音识别

1） 生成需要识别的wav文件，SpeechRecognition需要wav文件，不能识别mp3文件

安装库：

sudo apt install espeak ffmpeg libespeak1

pip install pyttsx3

代码：

def demo_tts_wav():

import pyttsx3

engine = pyttsx3.init()

engine.setProperty('rate', 150)

engine.setProperty('volume', 1.0)

voices = engine.getProperty('voices')

engine.setProperty('voice', voices0.id)

text = '你好，我是一个AI机器人'

#engine.say(text)

filename = 'ni_hao.wav'

engine.save_to_file(text, filename)

engine.runAndWait()

2. 语音识别，使用speech_recognition

安装库：

pip install SpeechRecognition

pip install pocketsphinx

下载模型文件：CMU Sphinx - Browse /Acoustic and Language Models/Mandarin at SourceForge.net

pip install vosk

下载模型文件到代码目录下：VOSK Models

解压，并且重命名为model

代码

def demo_speech_recognition():

import speech_recognition as sr

r = sr.Recognizer()

try:

audio_file = sr.AudioFile('ni_hao.wav')

with audio_file as source:

audio_data = r.record(source)

#text = r.recognize_google(audio_data, language='zh-Cn')

#text = r.recognize_wit(audio_data)

text = r.recognize_vosk(audio_data, language='zh-Cn')

print("识别结果：", text)

except Exception as e:

print("无法识别语音：", str(e))

3. 使用whisper库，效果最好，可以离线

安装：

pip install -U openai-whisper

权重文件不方便下载的话可以到这下载：https://download.csdn.net/download/love_xunmeng/88651611

然后移动到：

mv small.pt /home/user_account/.cache/whisper/

代码：

def demo_whisper():

import whisper

model = whisper.load_model("small")

result = model.transcribe("ni_hao.wav")

print(result"text")