Python3 通过轮询方式使用腾讯云语音识别接口实现录音文件转写

腾讯云最先是只支持使用回调的方式获取录音识别的结果的，当时我也针对回调的方式写过一篇博客https://blog.csdn.net/TomorrowAndTuture/article/details/100100430。但 9月5号左右有一次更新，腾讯针对录音文件的识别变成可支持轮询的方式获取识别结果，详细修改情况请查看腾讯云GitHub的Python源码，https://github.com/tencentcloud/tencentcloud-sdk-python：

原来的回调方式获取结果需要自己搭建一个服务，专门接收腾讯识别结果的POST请求，但是，如果用户没有官网可访问的IP或者域名的话，搭建好接收腾讯回调POST请求的地址并不算简单。现在支持通过TaskId获取识别结果的话，自然是再好不过的了，而且只要在一定期限之内，通过TaskId还可以重复获取识别结果。

首先，安装 tencentcloud-sdk-python 的包，你可以通过pip或者其他方式进行安装：

然后的话，我也不绕弯子了，直接上代码，先是根据需要上传录音文件url和相应的参数：

"""
@file: recognition_request.py
@author: Looking
@email: 2392863668@qq.com
"""from tencentcloud.common import credential
from tencentcloud.common.profile.client_profile import ClientProfile
from tencentcloud.common.profile.http_profile import HttpProfile
from tencentcloud.common.exception.tencent_cloud_sdk_exception import TencentCloudSDKException
from tencentcloud.asr.v20190614 import asr_client, modelstry:cred = credential.Credential("your SECRETID", "your SECRET_KEY")httpProfile = HttpProfile()httpProfile.endpoint = "asr.tencentcloudapi.com"clientProfile = ClientProfile()clientProfile.httpProfile = httpProfileclient = asr_client.AsrClient(cred, "ap-guangzhou", clientProfile)req = models.CreateRecTaskRequest()# 下面这个参数你自己根据需要进行设置params = '{"EngineModelType":"8k_6","ChannelNum":1,"ResTextFormat":0,"SourceType":0,"Url":"http://audio.c.---.wav"}'req.from_json_string(params)resp = client.CreateRecTask(req)print(resp.to_json_string())except TencentCloudSDKException as err:print(err)

然后运行可以得到类似如下的结果：

"D:\Program Files\Python36\python3.exe" D:/MyProject/Python/Voice_SDK/python_record_asr_sdk/src/request_recognition.py
{"Data": {"TaskId": 537731632}, "RequestId": "4d83b186-98fc-4d51-ba09-da65fe7b891e"}Process finished with exit code 0

这样的话，你就可以获取这通录音识别的 TaskId 了，接着通过 TaskId 获取录音识别结果，我这儿添加了部分对录音识别结果进行解析的代码：

"""
@file: recognition_result.py
@author: Looking
@email: 2392863668@qq.com
"""
from tencentcloud.common import credential
from tencentcloud.common.profile.client_profile import ClientProfile
from tencentcloud.common.profile.http_profile import HttpProfile
from tencentcloud.common.exception.tencent_cloud_sdk_exception import TencentCloudSDKException
from tencentcloud.asr.v20190614 import asr_client, models
try:cred = credential.Credential("your SECRETID", "your SECRET_KEY")httpProfile = HttpProfile()httpProfile.endpoint = "asr.tencentcloudapi.com"clientProfile = ClientProfile()clientProfile.httpProfile = httpProfileclient = asr_client.AsrClient(cred, "ap-guangzhou", clientProfile)req = models.DescribeTaskStatusRequest()params = '{"TaskId":537731632}'req.from_json_string(params)resp = client.DescribeTaskStatus(req)# print(resp.to_json_string())recognition_text = resp.to_json_string()# print(eval(recognition_text)['Data'])recognition_text = eval(recognition_text)['Data']['Result']doc = open('result.txt', 'w', encoding='utf-8')sentence_list = recognition_text.split('\n')[0:-1]  # 列表最后一个元素是空字符串for sentence in sentence_list:content = sentence.split('  ')[1]  # 获取单句通话内容begin_time = sentence.split('  ')[0].split(',')[0][1:]  # 获取每句话的开始时间begin_time = str(int(begin_time.split(":")[0]) * 60000 + int(begin_time.split(":")[1].replace(".", "")))end_time = sentence.split('  ')[0].split(',')[1]  # 获取每句话的结束时间end_time = str(int(end_time.split(":")[0]) * 60000 + int(end_time.split(":")[1].replace(".", "")))speaker = sentence.split('  ')[0].split(',')[-1][:-1]  # 获取说话人print(speaker + "\t" + content + '\t' + begin_time + '\t' + end_time)print(speaker + "\t" + content + '\t' + begin_time + '\t' + end_time, file=doc)doc.close()except TencentCloudSDKException as err:print(err)

然后对返回的字符串结果进行解析，获取完整录音识别结果：

还有一点，识别结果返回是以string的形式返回的，记得用 eval 函数转换成 Python 中的字典再进行解析。（还是用json.loads() 吧，返回语句 return resp.to_json_string() 返回的反正是json串，用 eval 可能会有 bug。）