精易论坛

标题: C#百度语音识别 [打印本页]

作者: 197076131 时间: 2014-10-27 15:51
标题: C#百度语音识别

<div class="blockcode"><blockquote>private BaiDuSDK baidu = new BaiDuSDK(“你的百度API Key”,“你的百度Secret Key”);
txtresult.Text = baidu.WavToText(file, "zh");

复制代码

using System;
using 备份器.DLL;
using Newtonsoft.Json.Linq;
using System.IO;
using Newtonsoft.Json;

namespace 百度开放测试
{
public class BaiDuSDK
{
      private HttpDLL HTTP = new HttpDLL();
      private string token = "";
      /// <summary>
      /// 初始化
      /// </summary>
      /// <param name="APIKey">百度API Key</param>
      /// <param name="SecretKey">百度Secret Key</param>
      public BaiDuSDK(string APIKey, string SecretKey)
      {
         GetToken( APIKey,SecretKey);

      }
      /// <summary>
      /// Wav识别，百度传进来的文件有格式要求，格式不正常通常识别不正确比如：多个不一样文件都是返回一个“的”字
      /// 注意你的音频文件的采样率别搞错，目前百度只支持8000，16000的采样率
      /// </summary>
      /// <param name="filename">文件名称</param>
      /// <param name="rate">8000,16000采样率</param>
      /// <param name="lan">中英标识（zh中文 , en英文）</param>
      /// <returns></returns>
      public string WavToText(string filename , string lan)
      {
         if (File.Exists(filename))
         {
               WavInfo wav =GetWavInfo(filename);
            int rate = int.Parse(wav.dwsamplespersec.ToString());
            //非16bit 位深的直接返回错误
            if (int.Parse(wav.wbitspersample.ToString()) != 16) return "错误：原始语音的录音格式目前只支持评测 8k/16k 采样率 16bit 位深的单声道语音，请自行转换";
            int cuid =11111;//这个随便填
            Byte[] byt = File.ReadAllBytes(filename);
            int len = byt.Length;
            string base64String = Convert.ToBase64String(byt);
            Customer cust = new Customer("wav", rate, 1, token, cuid, len, lan, base64String);
            string jsonStr = JsonConvert.SerializeObject(cust);  //序列化成 Json 格式
            string html = HTTP.Post("http://vop.baidu.com/server_api", jsonStr, "", "https://openapi.baidu.com/");
            try
            {
                  JObject jsons = JObject.Parse(html);
                  if (jsons["err_msg"].Value<string>() == "success.")
                  {

                     return jsons["result"][0].ToString();
                  }
                  else
                  {
                     return jsons["err_msg"].Value<string>();
                  }
            }
            catch (Exception)
            {
                  //异常提示。表示返回的的非JSON数据，意味着参数错误或不支持格式
                  return html;
            }

         }
         else
         {
            return "文件不存在";
         }
      }
      /// <summary>
      /// 取出access_token
      /// </summary>
      private void GetToken(string APIKey,string SecretKey)
      {

         string s = HTTP.GetHtml("https://openapi.baidu.com/oauth/2.0/token?grant_type=client_credentials&client_id=" + APIKey + "&client_secret=" + SecretKey,
            "",
            "https://openapi.baidu.com/");
         JObject json = JObject.Parse(s);
         token = json["access_token"].Value<string>();

      }
      public struct WavInfo
      {
         public string groupid;
         public string rifftype;
         public long filesize;
         public string chunkid;
         public long chunksize;
         public short wformattag; //记录着此声音的格式代号，例如WAVE_FORMAT_PCM，WAVE_F0RAM_ADPCM等等。
         public ushort wchannels; //记录声音的频道数。
         public ulong dwsamplespersec;//记录每秒采样率。 16000
         public ulong dwavgbytespersec;//记录每秒的数据量。
         public ushort wblockalign;//记录区块的对齐单位。
         public ushort wbitspersample;//记录每个取样所需的位元数。位深16
         public string datachunkid;
         public long datasize;
      }
      /// <summary>
      /// 取出WAV头信息
      /// </summary>
      /// <param name="strpath"></param>
      /// <returns></returns>
      private  static WavInfo GetWavInfo(string strpath)
      {
         WavInfo wavInfo = new WavInfo();
         FileInfo fi = new FileInfo(strpath);
         using (System.IO.FileStream fs = fi.OpenRead())
         {
            if (fs.Length >= 44)
            {
                  byte[] bInfo = new byte[44];
                  fs.Read(bInfo, 0, 44);
                  System.Text.Encoding.Default.GetString(bInfo, 0, 4);
                  if (System.Text.Encoding.Default.GetString(bInfo, 0, 4) == "RIFF" && System.Text.Encoding.Default.GetString(bInfo, 8, 4) == "WAVE" && System.Text.Encoding.Default.GetString(bInfo, 12, 4) == "fmt ")
                  {
                     wavInfo.groupid = System.Text.Encoding.Default.GetString(bInfo, 0, 4);
                     System.BitConverter.ToInt32(bInfo, 4);
                     wavInfo.filesize = System.BitConverter.ToInt32(bInfo, 4);
                     //wavInfo.filesize = Convert.ToInt64(System.Text.Encoding.Default.GetString(bInfo,4,4));
                     wavInfo.rifftype = System.Text.Encoding.Default.GetString(bInfo, 8, 4);
                     wavInfo.chunkid = System.Text.Encoding.Default.GetString(bInfo, 12, 4);
                     wavInfo.chunksize = System.BitConverter.ToInt32(bInfo, 16);
                     wavInfo.wformattag = System.BitConverter.ToInt16(bInfo, 20);
                     wavInfo.wchannels = System.BitConverter.ToUInt16(bInfo, 22);
                     wavInfo.dwsamplespersec = System.BitConverter.ToUInt32(bInfo, 24);
                     wavInfo.dwavgbytespersec = System.BitConverter.ToUInt32(bInfo, 28);
                     wavInfo.wblockalign = System.BitConverter.ToUInt16(bInfo, 32);
                     wavInfo.wbitspersample = System.BitConverter.ToUInt16(bInfo, 34);
                     wavInfo.datachunkid = System.Text.Encoding.Default.GetString(bInfo, 36, 4);
                     wavInfo.datasize = System.BitConverter.ToInt32(bInfo, 40);
                  }
            }
         }
         return wavInfo;

      }

}
/// <summary>
/// 定义JSON数据的一个类
/// </summary>
public class Customer
{
      public string format { get; set; }
      public int rate { get; set; }
      public int channel { get; set; }
      public string token { get; set; }
      public int len { get; set; }
      public int cuid { get; set; }
      public string lan { get; set; }
      public string speech { get; set; }
      public Customer()
      {

      }
      /// <summary>
      /// json参数  百度还有其它参数看需求自己加
      /// </summary>
      /// <param name="format">文件类型</param>
      /// <param name="rate">采样率</param>
      /// <param name="channel">声道</param>
      /// <param name="token">access_token 帮助文档说有效期一个月。所以没必要多次申请</param>
      /// <param name="cuid">用户 id，推荐使用 mac 地址/手机IMEI 等类似参数实际上任何参数都可以</param>
      /// <param name="len">文件长度</param>
      /// <param name="lan">中英标识（zh中文 , en英文）</param>
      /// <param name="speech">文件的base64String编码</param>
      public Customer(string format, int rate, int channel, string token, int cuid, int len, string lan, string speech)
      {
         this.format = format;
         this.rate = rate;
         this.channel = channel;
         this.token = token;
         this.len = len;
         this.speech = speech;
         this.cuid = cuid;
         this.lan = lan;
      }
}
}

作者: 小落落 时间: 2014-10-27 16:03
学习了

作者: 刘精杰 时间: 2014-12-27 17:17
这个楼主是不是和了一个dll

欢迎光临精易论坛 (https://125.confly.eu.org/)