语音识别，文本转语音，语音转文本

 

1. 语音合成.
	百度AipSpeech：
	synthesis():
		1: 与合成语音的文本.
		2. 语言.
		3. 客户端类型，1 
		4.option{语音合成参数}: pit 声调  spd 语速 vol 音量 per cosplay 合成语音角色.
		
2. 语音识别
	百度 AipSpeech
		 asr():
			1. 音频文件流 
			2. 音频文件格式(pcm)
			3. 音频采样率(16000, 8k/16k)
			4. option{识别语种}： dev_pid :1537  包含简单英文的普通话识别. 
		
		录音文件基本上没有直接pcm格式，需要对音频文件进行转换.
		ffmpeg:FFmpeg公司.涉及audio 处理的基本上都用它了.
		通过os.system(FFmpeg_cmdstr)得到新的pcm文件.
		

 

1.  百度注册账号.

 

案例1 .文本转语音

from aip import AipSpeech
from aip import AipNlp
from uuid import uuid4
import setting
import os

APP_ID = "11562884"
API_KEY = "9iOLKP9VCo4nsEf3N8dcOUmT"
SECRET_KEY = "aW0kwOHFbHrQely6bcmGTzU49t2jOYdL"

baidu_client = AipSpeech(APP_ID, API_KEY, SECRET_KEY)
baidu_nlp = AipNlp(APP_ID, API_KEY, SECRET_KEY)

#.1 文本转语音.
def text2audio(text):

    file_name = f"{uuid4()}.mp3"
    print("filename:%s"%file_name)
    file_path = os.path.join(setting.AUDIO_FILE_PATH, file_name)
    print("file_path:%s" %(file_path))
    res = baidu_client.synthesis(text, "zh", 1, {
        "vol": 5,
        "pit": 7,
        "spd": 4,
        "per": 4
    })

    if type(res) == dict:
        return res

    with open(file_path, "wb") as f:
        f.write(res)

    return file_name

print(text2audio("强哥威武"))

打印结果：

filename:86375f56-f42d-447a-9bcc-829d370ff805.mp3
file_path:audio\86375f56-f42d-447a-9bcc-829d370ff805.mp3
86375f56-f42d-447a-9bcc-829d370ff805.mp3

 

案例2.语音转文本.

def audio2text(file_name):
    file_pcm_path =os.path.join(setting.AUDIO_FILE_PATH,file_name)
    cmd_str =f"ffmpeg -y -i {file_pcm_path} -acodec pcm_s16le -f s16le -ac 1 -ar 16000 {file_pcm_path}.pcm"
    os.system(cmd_str)

    with open(f"{file_pcm_path}","rb") as f :
        audio_context =f.read()

    res =baidu_client.asr(audio_context,"pcm",16000,{"dev_pid": 1537})

    if res.get('err_no'):
        return res

    return res.get("result")[0]

print(audio2text("qiang.wma"))

 

案例3. 语音识别 NLP 

#.3 语音识别
def my_nlp(text):
    if baidu_nlp.simnet("你今年几岁了 ",text).get("score")>=0.72:
        print(baidu_nlp.simnet("你今年几岁了 ",text).get("score"))
        return "我今年73岁了，不然84岁也行"

    if baidu_nlp.simnet("你叫什么名字",text).get("score")>=0.72:
        return "我的名字叫做小嘿嘿"

    if baidu_nlp.simnet("你在哪里学习",text).get("score")>=0.72:
        return "我在学习python"


print(my_nlp('你多大了'))  #结果：我今年73岁了，不然84岁也行

print(my_nlp("你的名字是")) #结果：我的名字叫做小嘿嘿

  

代码总结 ：

from aip import AipSpeech
from aip import AipNlp
from uuid import uuid4
import setting
import os

APP_ID = "11562884"
API_KEY = "9iOLKP9VCo4nsEf3N8dcOUmT"
SECRET_KEY = "aW0kwOHFbHrQely6bcmGTzU49t2jOYdL"

baidu_client = AipSpeech(APP_ID, API_KEY, SECRET_KEY)
baidu_nlp = AipNlp(APP_ID, API_KEY, SECRET_KEY)

#.1 文本转语音.
def text2audio(text):

    file_name = f"{uuid4()}.mp3"
    print("filename:%s"%file_name)
    file_path = os.path.join(setting.AUDIO_FILE_PATH, file_name)
    print("file_path:%s" %(file_path))
    res = baidu_client.synthesis(text, "zh", 1, {
        "vol": 5,
        "pit": 7,
        "spd": 4,
        "per": 4
    })

    if type(res) == dict:
        return res

    with open(file_path, "wb") as f:
        f.write(res)

    return file_name
#
# print(text2audio("强哥威武"))



#.2 语音转文本.
def audio2text(file_name):
    file_pcm_path =os.path.join(setting.AUDIO_FILE_PATH,file_name)
    cmd_str =f"ffmpeg -y -i {file_pcm_path} -acodec pcm_s16le -f s16le -ac 1 -ar 16000 {file_pcm_path}.pcm"
    os.system(cmd_str)

    with open(f"{file_pcm_path}","rb") as f :
        audio_context =f.read()

    res =baidu_client.asr(audio_context,"pcm",16000,{"dev_pid": 1537})

    if res.get('err_no'):
        return res

    return res.get("result")[0]
#
print(audio2text("qiang.wma"))


#.3 语音识别
def my_nlp(text):
    if baidu_nlp.simnet("你今年几岁了 ",text).get("score")>=0.72:
        print(baidu_nlp.simnet("你今年几岁了 ",text).get("score"))
        return "我今年73岁了，不然84岁也行"

    if baidu_nlp.simnet("你叫什么名字",text).get("score")>=0.72:
        return "我的名字叫做小嘿嘿"

    if baidu_nlp.simnet("你在哪里学习",text).get("score")>=0.72:
        return "我在学习python"


print(my_nlp('你多大了'))  #结果：我今年73岁了，不然84岁也行

print(my_nlp("你的名字是")) #结果：我的名字叫做小嘿嘿

 

 

send_file 与audio标签.

py文件

from  flask  import Flask,request,send_file
import setting
import os

app =Flask(__name__)



@app.route("/getfile/<file_name>")
def get_file(file_name):

    audio_file =os.path.join(setting.AUDIO_FILE_PATH,file_name)

    return send_file(audio_file)


if __name__ =="__main__":
    app.run()

 

html文件

<!DOCTYPE html>
<html lang="en">
<head>
    <meta charset="UTF-8">
    <title>Title</title>
</head>
<body>
<audio autoplay="autoplay" id="play_mp3" src="http://127.0.0.1:1111/getfile/7b86f60d-6f43-4342-9948-38657de4843e.mp3"></audio>
</body>
</html>