C# 中的 Google 语音转文本 API
Posted
技术标签:
【中文标题】C# 中的 Google 语音转文本 API【英文标题】：Google speech to text API in C# 【发布时间】：2014-07-25 19:34:54 【问题描述】：当flac文件使用windows的录音机手动录制并使用软件转换器进行转换时，我的程序得到了谷歌的正确响应。 但是当我使用我的程序记录的文件时，我得到了 ""result":[] “来自谷歌。我该怎么办？ 这是我的代码： 发件人：
private static void CopyStream(FileStream fileStream, Stream requestStream)
var buffer = new byte[32768];
int read;
while ((read = fileStream.Read(buffer, 0, buffer.Length)) > 0)
requestStream.Write(buffer, 0, read);
private static void ConfigureRequest(HttpWebRequest request)
request.KeepAlive = true;
request.SendChunked = true;
request.ContentType = "audio/x-flac; rate=44100";
request.UserAgent =
"Mozilla/5.0 (Windows NT 6.2; WOW64) AppleWebKit/535.2 (Khtml, like Gecko) Chrome/15.0.874.121 Safari/535.2";
request.Headers.Set(HttpRequestHeader.AcceptEncoding, "gzip,deflate,sdch");
request.Headers.Set(HttpRequestHeader.AcceptLanguage, "en-GB,en-US;q=0.8,en;q=0.6");
request.Headers.Set(HttpRequestHeader.AcceptCharset, "ISO-8859-1,utf-8;q=0.7,*;q=0.3");
request.Method = "POST";
using (var fileStream = new FileStream(@"C:\Users\Ahmad Mustofa\Documents\Visual Studio 2010\Projects\FP\FP\bin\Debug\voice.flac", FileMode.Open))
const string requestUrl = "https://www.google.com/speech-api/v2/recognize?output=json&lang=ar-sa&key=AIzaSyBJ6VJ326Rpb23msih2wGhXENEwU1TF1PA&client=chromium&maxresults=1&pfilter=2";
var request = (HttpWebRequest)WebRequest.Create(requestUrl);
ConfigureRequest(request);
var requestStream = request.GetRequestStream();
CopyStream(fileStream, requestStream);
using (var response = request.GetResponse())
using (var responseStream = response.GetResponseStream())
using (var zippedStream = new GZipStream(responseStream, CompressionMode.Decompress))
using (var sr = new StreamReader(zippedStream))
var res = sr.ReadToEnd();
state.Text = res;
wav 录音机：
private void sourceStream_DataAvailable(object sender, NAudio.Wave.WaveInEventArgs e)
if (waveWriter == null) return;
waveWriter.WriteData(e.Buffer, 0, e.BytesRecorded);
waveWriter.Flush();
fileName = "C:\\Users\\Ahmad Mustofa\\Documents\\Visual Studio 2010\\Projects\\FP\\FP\\bin\\debug\\voice.wav";
int deviceNumber = hardware.SelectedItems[0].Index;
try
sourceStream = new NAudio.Wave.WaveIn();
sourceStream.DeviceNumber = deviceNumber;
sourceStream.WaveFormat = new NAudio.Wave.WaveFormat(44100, NAudio.Wave.WaveIn.GetCapabilities(deviceNumber).Channels);
sourceStream.DataAvailable += new EventHandler<NAudio.Wave.WaveInEventArgs>(sourceStream_DataAvailable);
waveWriter = new NAudio.Wave.WaveFileWriter(fileName, sourceStream.WaveFormat);
sourceStream.StartRecording();
catch (Exception ex)
state.Text = "disini" + ex.Message;
flac 转换器：
string inputFile = Path.Combine("wav ", input);
string outputFile = Path.Combine("flac", Path.ChangeExtension(input, ".flac"));
if (!File.Exists(inputFile))
throw new ApplicationException("Input file " + inputFile + " cannot be found!");
WavReader wav = new WavReader(inputFile);
using (var flacStream = File.Create(outputFile))
FlacWriter flac = new FlacWriter(flacStream, wav.BitDepth, wav.Channels, wav.SampleRate);
// Buffer for 1 second's worth of audio data
byte[] buffer = new byte[wav.Bitrate / 8];
int bytesRead;
do
bytesRead = wav.InputStream.Read(buffer, 0, buffer.Length);
flac.Convert(buffer, 0, bytesRead);
while (bytesRead > 0);
flac.Dispose();
flac = null;
【问题讨论】：
我也有同样的问题 【参考方案1】：我也遇到了同样的问题，但想出了一个巧妙的解决方案。我使用 Fiddler (http://www.telerik.com/fiddler/) 弄清楚 Chrome 是如何进行语音识别的，然后创建了一些代码来模拟 chrome 发送请求。这种方法使用不同的 URI，并且还有一个 16 字符的值，称为对，每个请求都不同。我使用一个简单的随机值生成器函数来为请求创建一个，我还将输出值更改为“json”。
注意：结果有时可能为空，如上述情况，但响应中还有另一个包含替代项的 json 对象。
private void GoogleSpeechToText()
string uri = "https://www.google.com/speech-api/full-duplex/v1/up?output=json&key=AIzaSyBOti4mM-6x9WDnZIjIeyEU21OpBXqWBgw&pair=" + GenerateUnique(16) + "&lang=en-US&pFilter=2&maxAlternatives=10&client=chromium";
HttpWebRequest request = (HttpWebRequest)WebRequest.Create(uri);
request.Timeout = 10000;
request.Method = "POST";
request.Host = "www.google.com";
request.KeepAlive = true;
request.SendChunked = true;
request.ContentType = "audio/x-flac; rate=16000";
request.Headers.Set(HttpRequestHeader.AcceptLanguage, "en-GB,en-US;q=0.8,en;q=0.6");
request.UserAgent = "Mozilla/5.0 (Windows NT 6.3; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/36.0.1985.143 Safari/537.36";
string path = @"C:\TestFolder\test_audio.flac";
FileInfo fInfo = new FileInfo(path);
var numBytes = fInfo.Length;
byte[] data;
using (FileStream fStream = new FileStream(path, FileMode.Open, FileAccess.Read))
data = new Byte[numBytes];
fStream.Read(data, 0, (int) numBytes);
fStream.Close();
using (Stream reqStream = request.GetRequestStream())
reqStream.Write(data, 0, data.Length);
try
WebResponse response = request.GetResponse();
Stream respStream = response.GetResponseStream();
if(response.ContentType == "application/json; charset=utf-8")
using (var sr = new StreamReader(respStream))
var res = sr.ReadToEnd();
textBox1.Text = res;
catch (Exception ex) MessageBox.Show(ex.Message, "Error", MessageBoxButtons.OK);
private string GenerateUnique(int length)
string[] LETTERS = new string[] "A", "B", "C", "D", "E", "F", "G", "H", "I", "J", "K", "L", "M", "N", "O", "P", "Q", "R", "S", "T", "U", "V", "W", "X", "Y", "Z" ;
string[] DIGITS = new string[] "0", "1", "2", "3", "4", "5", "6", "7", "8", "9" ;
string buffer = "";
Random random = new Random();
for(int i = 0; i < length; i++)
int rnd = random.Next(2);
if (rnd == 1)
buffer += LETTERS[random.Next(LETTERS.Length)];
else
buffer += DIGITS[random.Next(DIGITS.Length)];
return buffer;
【讨论】：
【参考方案2】：Google cloud api url中明确提到了，即
https://cloud.google.com/speech-to-text/docs/async-recognize#speech-async-recognize-gcs-protocol
如果操作尚未完成，您可以通过重复发出 GET 请求来轮询端点，直到响应的 done 属性为 true。
"name": "operationname here",
"metadata":
"@type": "type.googleapis.com/google.cloud.speech.v1.LongRunningRecognizeMetadata",
"progressPercent": 0,
"startTime": "2018-12-18T10:56:09.425584Z",
"lastUpdateTime": "2018-12-18T11:10:27.147310Z"
,
"done": true,
通过重复发出 GET 请求来轮询端点，直到响应的 done 属性为真，或者您可以检查“progressPercent”：0 直到它的值变为 100。一旦它的 100% 则 done 属性变为真。
我在我的代码中使用操作名称做了同样的事情，这里是代码供参考
public async Task<string> TranscribeLongMediaFile(string operationName)
string bearerToken = GetOAuthToken();
var baseUrl = new Uri(googleSpeechBaseUrl + operationName);
string resultContent = string.Empty;
using (var client = new HttpClient())
client.DefaultRequestHeaders.Add(HttpRequestHeader.Authorization.ToString(), "Bearer " + bearerToken);
client.DefaultRequestHeaders.Add(HttpRequestHeader.ContentType.ToString(), "application/json; charset=utf-8");
client.Timeout = TimeSpan.FromMilliseconds(Timeout.Infinite);
int currentPercentage = 0;
bool responseStatus = false;
while (!responseStatus)
responseStatus = false;
// Send request
using (var result = await client.GetAsync(baseUrl))
resultContent = await result.Content.ReadAsStringAsync();
ResponseObject responseObject = JsonConvert.DeserializeObject<ResponseObject>(resultContent);
currentPercentage = responseObject.metadata.progressPercent;
responseStatus = (responseObject.done && currentPercentage == 100);
// Delay the request based on percentage value to repeatedly making the GET request until the done property of the response is true.
await Task.Delay(CalculateDealy(currentPercentage));
;
return resultContent;
为了延迟get请求：
/// <summary>
/// Delay the request to number of milliseconds
/// </summary>
/// <param name="currentPercentage"></param>
/// <returns></returns>
private int CalculateDealy(int currentPercentage)
int x = currentPercentage / 10;
return (10 - x) * 1500;
获取授权令牌：
/// <summary>
/// Get OAuth token
/// </summary>
/// <returns></returns>
public string GetOAuthToken()
return googleCredential.UnderlyingCredential.GetAccessTokenForRequestAsync("https://accounts.google.com/o/oauth2/v2/auth", CancellationToken.None).Result;
最后，你会得到如下结果：
"name": "operationname here",
"metadata":
"@type": "type.googleapis.com/google.cloud.speech.v1.LongRunningRecognizeMetadata",
"progressPercent": 100,
"startTime": "2018-12-18T10:56:09.425584Z",
"lastUpdateTime": "2018-12-18T11:10:27.147310Z"
,
"done": true,
"response":
"@type": "type.googleapis.com/google.cloud.speech.v1.LongRunningRecognizeResponse",
"results": [
"alternatives": [
"transcript": "okay let's get started",
"confidence": 0.97442055
]
, and so on .....
需要的东西：
-
api-key.json 文件
安装包 Google.Apis.Auth.OAuth2 以授权
HTTP 网络请求
谢谢
【讨论】：
【参考方案3】：使用它并获得话语（短语）和置信度（％）
string toParse=(VALUE RETURNED BY GOOGLE)
var trsc1 = @"transcript"":""";
var trsc2 = @""",""confidence"":";
var trsc3 = @"],""final"":";
var start = toParse.IndexOf(trsc1) + trsc1.Length;
var end = toParse.IndexOf(trsc2);
var end2 = toParse.IndexOf(trsc3);
var vv1 = toParse.Substring(start, end - start);
var vv2 = toParse.Substring(end + trsc2.Length, end2 - (end + trsc2.Length));
vv2 = vv2.Trim().Replace(".", ",");
float confidence = (float)Math.Round(double.Parse(vv2), 2);
string utterance = vv1;
【讨论】：
【参考方案4】：using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.IO;
using System.Net;
using Newtonsoft.Json;
namespace google_speech_api_trial4
class Program
public static string ACCESS_GOOGLE_SPEECH_KEY = "AIzaSyDC8nM1S0cLpXvRc8TXrDoey-tqQsoBGnM";
static void Main(string[] args)
GoogleSpeechRequest();
Console.ReadLine();
public static void GoogleSpeechRequest()
FileStream fileStream = File.OpenRead("my.flac");
MemoryStream memoryStream = new MemoryStream();
memoryStream.SetLength(fileStream.Length);
fileStream.Read(memoryStream.GetBuffer(), 0, (int)fileStream.Length);
byte[] BA_AudioFile = memoryStream.GetBuffer();
HttpWebRequest _HWR_SpeechToText = null;
_HWR_SpeechToText = (HttpWebRequest)HttpWebRequest.Create("https://www.google.com/speech-api/v2/recognize?output=json&lang=en-us&key=" + ACCESS_GOOGLE_SPEECH_KEY);
_HWR_SpeechToText.Credentials = CredentialCache.DefaultCredentials;
_HWR_SpeechToText.Method = "POST";
_HWR_SpeechToText.ContentType = "audio/x-flac; rate=44100";
_HWR_SpeechToText.ContentLength = BA_AudioFile.Length;
Stream stream = _HWR_SpeechToText.GetRequestStream();
stream.Write(BA_AudioFile, 0, BA_AudioFile.Length);
stream.Close();
HttpWebResponse HWR_Response = (HttpWebResponse)_HWR_SpeechToText.GetResponse();
StreamReader SR_Response = new StreamReader(HWR_Response.GetResponseStream());
string responseFromServer = (SR_Response.ReadToEnd());
String[] jsons = responseFromServer.Split('\n');
String text = "";
foreach (String j in jsons)
dynamic jsonObject = JsonConvert.DeserializeObject(j);
if (jsonObject == null || jsonObject.result.Count <= 0)
continue;
text = jsonObject.result[0].alternative[0].transcript;
Console.WriteLine(text);
我找了 3 个小时，当我打印文本时，我一直在寻找 "result : []"。我以为它没有转换音频。但是，Json obj 有两行。第二行有音频文本。要打印它，我们需要解析它。哦，我在导入、引用和使用语句方面也遇到了很多问题。 但最后这段代码有效。
【讨论】：
以上是关于C# 中的 Google 语音转文本 API的主要内容，如果未能解决你的问题，请参考以下文章
Android TextToSpeech TTS中文文本转语音(语音合成)