由於項目需要,這幾天都在試圖利用百度語音API進行語音識別。但是識別到的都是“啊,哦”什么的,我就哭了。
這里我只是分享一下這個過程,錯誤感覺出現在Post語音數據那一塊,可能是轉換問題吧。
API請求地址::http://vop.baidu.com/server_api
語音上傳模式:顯示發送:將語音數據直接放在 HTTP-BODY 中
其他參數:cuid:用戶id,token:密鑰 ,lan:語言等
要了解更多請查看官方文檔:http://developer.baidu.com/wiki/index.php?title=docs/cplat/media/voice
實現的步驟:1、錄音,將文件存儲起來 2、獲取token 3、Post數據並請求獲取返回值
1、錄音
String fileName = "test.wav"; private MediaCapture _mediaCaptureManager; private StorageFile _recordStorageFile; private async void record_Click(object sender, RoutedEventArgs e)//點擊后開始錄音 { try { //在臨時文件夾Temp中創建文件,存在的話就替換掉 _recordStorageFile = await ApplicationData.Current.TemporaryFolder.CreateFileAsync(fileName, CreationCollisionOption.ReplaceExisting); //關鍵就是這倆句 MediaEncodingProfile recordProfile = MediaEncodingProfile.CreateWav(AudioEncodingQuality.Auto);//錄音-WAV格式 await _mediaCaptureManager.StartRecordToStorageFileAsync(recordProfile, this._recordStorageFile);//將錄音保存到創建的文件中 } catch (Exception ex) { Debug.WriteLine(ex.Message.ToString()); } } private async void stop_Click(object sender, RoutedEventArgs e) { await _mediaCaptureManager.StopRecordAsync();//停止錄音 }
2、獲取token
private void GetToken() { HttpWebRequest request = (HttpWebRequest)WebRequest.Create("https://openapi.baidu.com/oauth/2.0/token?grant_type=client_credentials&client_id=" + "申請的API_KEY" + "&client_secret=" + “申請的SECRET_KEY”); request.BeginGetResponse(ResponseTokenCall, request); } private void ResponseTokenCall(IAsyncResult result) { try { HttpWebRequest httpWebRequest = (HttpWebRequest)result.AsyncState; WebResponse webResponse = httpWebRequest.EndGetResponse(result); using (Stream stream = webResponse.GetResponseStream()) using (StreamReader reader = new StreamReader(stream)) { string content = reader.ReadToEnd(); string ssss = content.Replace("\"", "").Replace("{", "").Replace("}", "").Replace("\n", ""); string[] indexs = ssss.Split(','); foreach (string index in indexs) { string[] _indexs = index.Split(':'); if (_indexs[0] == "access_token") token = _indexs[1];//獲取到的token } } } catch (Exception ex) { Debug.WriteLine("獲取Token失敗"); } }
3、發送請求
private async void Post() { string serverURL = "http://vop.baidu.com/server_api?"; HttpWebRequest request = (HttpWebRequest)WebRequest.Create(serverURL + "lan=zh&cuid=" + cuid + "&token=" + token);//cuid = Guid.NewGuid().ToString(); 官方推薦使用 mac 地址/手機 IMEI 等類似參數,只要唯一就好 request.ContinueTimeout = 10000;//超時 request.Method = "POST";//Post請求 //post data request.BeginGetRequestStream(ResponseStreamCallbackPost, request);//Post數據 } private async void ResponseStreamCallbackPost(IAsyncResult result) { StorageFile storageFile = await ApplicationData.Current.TemporaryFolder.GetFileAsync(fileName);//取出臨時文件夾中保存的音頻文件 IBuffer buffer = await FileIO.ReadBufferAsync(storageFile);//讀取文件至Ibuffer byte[] voice = WindowsRuntimeBufferExtensions.ToArray(buffer, 0, (int)buffer.Length);//將Ibuffer轉換為byte[]
HttpWebRequest httpWebRequest = (HttpWebRequest)result.AsyncState; httpWebRequest.ContentType = "audio/wav;rate=8000";//參數設置 using (Stream writeStream = httpWebRequest.EndGetRequestStream(result)) { writeStream.Write(voice, 0, voice.Length);//寫入 } httpWebRequest.BeginGetResponse(ResponseCall, httpWebRequest);//發送請求 } private void ResponseCall(IAsyncResult result) { try { HttpWebRequest httpWebRequest = (HttpWebRequest)result.AsyncState; WebResponse webResponse = httpWebRequest.EndGetResponse(result); using (Stream stream = webResponse.GetResponseStream()) using (StreamReader reader = new StreamReader(stream)) { string content = reader.ReadToEnd();//返回的是utf-8編碼 string a = Regex.Unescape(content);//進行轉換 } } catch (Exception ex) { Debug.WriteLine("識別失敗"); } }
token獲取了一次,是說30天后會過期。
在處理過程中,遇到了一個問題。我讓GetToken()和Post()先后進行處理,但是卻發現token是在Post之后才得到的,這就導致了請求的URL里面不含token了。
所以還是要先獲取了token之后再進行。
另外(HttpWebRequest)request無法給Content-length屬性進行賦值。