科大讯飞语音转写API

科大讯飞语音转写API

一、语音转写是什么？

语音转写（Long Form ASR）基于深度全序列卷积神经网络，将长段音频（5小时以内）数据转换成文本数据，为信息处理和数据挖掘提供基础。转写的是已录制音频（非实时），音频文件上传成功后进入等待队列，待转写成功后用户即可获取结果，返回结果时间受音频时长以及排队任务量的影响。 如遇转写耗时比平时延长，大概率表示当前时间段出现转写高峰，请耐心等待即可。

二、使用步骤

1.注册并认证

访问讯飞官网注册账号：https://xinghuo.xfyun.cn/sparkapi?ch=gji

服务管理-完成个人认证

2.创建应用

右上角-我的应用

完善应用信息

左边选择“语音识别”-“语音转写”

右边的token信息需要记录下来
免费识别时长为5小时

3.构建API

https://www.xfyun.cn/doc/asr/ifasr_new/API.html
官方提供的demo有python3、java语言，这里以PHP调用python为例构建一个API
index.php获取待识别的音频，存储到本地供Python调用，识别完成后删除音频。

<?php
header('Access-Control-Allow-Origin:*');
header('Content-type: application/json');
 
$url=isset($_GET['url'])? $_GET['url'] :null; 
if(empty($url)){die("请传入音频链接参数");}
 
preg_match('/[^.]+$/', $url, $matches);
$extension = $matches[0];
$array = ["mp3","wav","pcm","aac","opus","flac","ogg","m4a","amr","speex","lyb","ac3","aac","ape","m4r","mp4","acc","wma"];
if(!in_array($extension, $array)){die("当前音频格式不支持");}
$file_path = dirname(__FILE__) . '/cache/' . time() . '.'.$extension;
$file_data = file_get_contents($url); // 从URL获取文件数据
file_put_contents($file_path, $file_data); // 将文件数据保存到本地
 
$str = exec("python3 lfasr-new.py $file_path");
// 将处理后的JSON字符串转换为PHP对象或数组
$data = json_decode($str, true); 
 
$response_msg = $data["code"];
if($response_msg == 000000) {$code = "200";}
else{$code = "202";}
 
$res = $data["content"]["orderResult"];
$res = json_decode($res, true); 
$res = $res["lattice2"];
$num = count($res);
for ($i=0; $i<$num; $i++)
{
    $slice = $res[$i]["json_1best"]["st"]["rt"][0]["ws"];
    $totalItems = count($slice);
    for ($j=0; $j<$totalItems; $j++)
    {
    $words .= $slice[$j]["cw"][0]["w"];
    }
}
 
$json_return = array(
    "code" => $code,
    "src" => $url,
    "dst" => $words
);
echo json_encode($json_return, JSON_UNESCAPED_UNICODE);
unlink($file_path);

lfasr-new.py文件写入内容，请求讯飞识别接口

# -*- coding: utf-8 -*-
import base64
import hashlib
import hmac
import json
import os
import time
import requests
import urllib
import sys
 
lfasr_host = 'https://raasr.xfyun.cn/v2/api'
# 请求的接口名
api_upload = '/upload'
api_get_result = '/getResult'
 
 
class RequestApi(object):
    def __init__(self, appid, secret_key, upload_file_path):
        self.appid = appid
        self.secret_key = secret_key
        self.upload_file_path = upload_file_path
        self.ts = str(int(time.time()))
        self.signa = self.get_signa()
 
    def get_signa(self):
        appid = self.appid
        secret_key = self.secret_key
        m2 = hashlib.md5()
        m2.update((appid + self.ts).encode('utf-8'))
        md5 = m2.hexdigest()
        md5 = bytes(md5, encoding='utf-8')
        # 以secret_key为key, 上面的md5为msg， 使用hashlib.sha1加密结果为signa
        signa = hmac.new(secret_key.encode('utf-8'), md5, hashlib.sha1).digest()
        signa = base64.b64encode(signa)
        signa = str(signa, 'utf-8')
        return signa
 
 
    def upload(self):
        # print("上传部分：")
        upload_file_path = self.upload_file_path
        file_len = os.path.getsize(upload_file_path)
        file_name = os.path.basename(upload_file_path)
 
        param_dict = {}
        param_dict['appId'] = self.appid
        param_dict['signa'] = self.signa
        param_dict['ts'] = self.ts
        param_dict["fileSize"] = file_len
        param_dict["fileName"] = file_name
        param_dict["duration"] = "200"
        # print("upload参数：", param_dict)
        data = open(upload_file_path, 'rb').read(file_len)
 
        response = requests.post(url =lfasr_host + api_upload+"?"+urllib.parse.urlencode(param_dict),
                                headers = {"Content-type":"application/json"},data=data)
        # print("upload_url:",response.request.url)
        result = json.loads(response.text)
        # print("upload resp:", result)
        return result
 
 
    def get_result(self):
        uploadresp = self.upload()
        orderId = uploadresp['content']['orderId']
        param_dict = {}
        param_dict['appId'] = self.appid
        param_dict['signa'] = self.signa
        param_dict['ts'] = self.ts
        param_dict['orderId'] = orderId
        param_dict['resultType'] = "transfer,predict"
        # print("")
        # print("查询部分：")
        # print("get result参数：", param_dict)
        status = 3
        # 建议使用回调的方式查询结果，查询接口有请求频率限制
        while status == 3:
            response = requests.post(url=lfasr_host + api_get_result + "?" + urllib.parse.urlencode(param_dict),
                                     headers={"Content-type": "application/json"})
            # print("get_result_url:",response.request.url)
            result = json.loads(response.text)
            # print(result)
            status = result['content']['orderInfo']['status']
            # print("status=",status)
            if status == 4:
                break
            time.sleep(1)
        print(json.dumps(result))
        return result
 
# 输入讯飞开放平台的appid，secret_key和待转写的文件路径
if __name__ == '__main__':
    api = RequestApi(appid="xxxxxx",
                     secret_key="xxxxxx",
                     upload_file_path=sys.argv[1])
 
    api.get_result()

运行前：请先填写Appid、SecretKey，新建cache缓存文件夹。

4.调用实例

请求地址：

https://你的api地址/?url=待识别的音频文件网络地址

返回数据：

{
   "code": "200",
   "src": "待识别的音频文件网络地址",
   "dst": "识别后的文字"
}

Demo：https://api.szfx.top/api/lfasr.html

---

总结

以上就是使用PHP调用python构建语音识别 API 的教程，欢迎大家去体验。